1
00:00:01,760 --> 00:00:03,679
Join us as we gather around the hedge,

2
00:00:03,679 --> 00:00:06,660
where we dig into technology, business, and culture

3
00:00:06,719 --> 00:00:09,059
with the finest minds in computer networking.

4
00:00:21,054 --> 00:00:22,035
Well, good morning,

5
00:00:22,734 --> 00:00:23,234
Audrey.

6
00:00:24,974 --> 00:00:26,195
Audrey says hi.

7
00:00:27,375 --> 00:00:28,594
Hi. Hi, Tom.

8
00:00:28,969 --> 00:00:30,969
Oh, hi. Oh, yeah. I'm here too. I'm

9
00:00:30,969 --> 00:00:32,890
glad to see that you joined Audrey this

10
00:00:32,890 --> 00:00:34,590
morning for this episode of The Hedge.

11
00:00:34,969 --> 00:00:36,030
Audrey's the plant.

12
00:00:36,409 --> 00:00:38,649
Got it. I thought I remembered the plant

13
00:00:38,649 --> 00:00:41,049
reference. Yeah. Yeah. I have I have virtual

14
00:00:41,049 --> 00:00:42,925
plants, so I came prepared. I don't have

15
00:00:42,925 --> 00:00:44,945
I don't have names, though. So sorry.

16
00:00:45,965 --> 00:00:46,784
That's okay.

17
00:00:47,405 --> 00:00:49,804
And this morning, we're joined by Scott Robyn.

18
00:00:49,804 --> 00:00:51,344
By the way, Tom, are you warm?

19
00:00:52,765 --> 00:00:54,844
Somewhat. I mean, it's gonna be, like, 65

20
00:00:54,844 --> 00:00:55,405
today. So

21
00:00:56,210 --> 00:00:58,130
Oh. Oh, well, no. I don't know. Just

22
00:00:58,130 --> 00:01:00,129
trying to make everybody jealous who's there. It

23
00:01:00,129 --> 00:01:02,629
works. You just did. You just did.

24
00:01:03,729 --> 00:01:05,569
Scott, where are you? I don't remember where

25
00:01:05,569 --> 00:01:07,269
you are. I am in Harrisonburg,

26
00:01:07,569 --> 00:01:10,129
Virginia in the beautiful Shenandoah Valley, not that

27
00:01:10,129 --> 00:01:12,634
far from you. So Yeah. Not that far.

28
00:01:12,935 --> 00:01:14,935
I think it's still a bit colder and

29
00:01:14,935 --> 00:01:17,415
snowier there than it is for me. Maybe.

30
00:01:17,415 --> 00:01:19,334
I don't know. We're we're we're going below

31
00:01:19,334 --> 00:01:21,575
zero at night here still. So Yeah. Well,

32
00:01:21,575 --> 00:01:23,575
we are too. I mean, we do it.

33
00:01:23,575 --> 00:01:25,274
Yeah. We still do in in,

34
00:01:25,894 --> 00:01:28,890
in Knoxville. It's still pretty darn cold in

35
00:01:28,890 --> 00:01:31,290
Knoxville, and they aren't planning on some snow

36
00:01:31,290 --> 00:01:33,229
this weekend because they brined the roads.

37
00:01:34,090 --> 00:01:36,989
Like, whatever. Nobody's gonna drive anyway.

38
00:01:38,170 --> 00:01:40,984
Like, we're all staying home. Although, I may

39
00:01:40,984 --> 00:01:42,265
go try to go to the gym on

40
00:01:42,265 --> 00:01:44,024
Sunday even though it snows. We'll see if

41
00:01:44,024 --> 00:01:45,084
the gym is open.

42
00:01:45,625 --> 00:01:48,604
That'll be the challenge. Will the gym open?

43
00:01:49,384 --> 00:01:52,099
I I doubt it. So today, we are

44
00:01:52,099 --> 00:01:54,920
talking about MPLS, not the city,

45
00:01:57,939 --> 00:01:58,680
the protocol.

46
00:01:59,379 --> 00:02:00,599
And there are many,

47
00:02:01,140 --> 00:02:03,944
many ways to say this, mostly pointless label

48
00:02:04,025 --> 00:02:06,045
switching. What did you say, Tom? Mipples?

49
00:02:06,424 --> 00:02:06,924
Mipples?

50
00:02:08,824 --> 00:02:10,685
It's a little cringey, but, you know,

51
00:02:14,185 --> 00:02:16,739
I can tell you're entertained, Russ. I never

52
00:02:16,739 --> 00:02:18,900
had a problem with MPLS. Like, it's not

53
00:02:18,900 --> 00:02:19,799
too many syllables.

54
00:02:20,099 --> 00:02:21,780
Right? I never felt the need to shorten

55
00:02:21,780 --> 00:02:23,079
that. Four syllables.

56
00:02:23,379 --> 00:02:23,879
Yeah.

57
00:02:24,419 --> 00:02:25,079
I mean,

58
00:02:25,379 --> 00:02:26,840
you know, it's kind of counterintuitive

59
00:02:27,219 --> 00:02:29,719
because the entire point of MPLS

60
00:02:30,099 --> 00:02:31,400
is to have a shortened

61
00:02:32,105 --> 00:02:32,605
forwarding

62
00:02:33,224 --> 00:02:33,724
paradigm.

63
00:02:34,105 --> 00:02:36,205
Do not have to forward on the entire

64
00:02:36,905 --> 00:02:38,745
IP v six address or IP v four

65
00:02:38,745 --> 00:02:40,665
address, but there's, like, this little short little

66
00:02:40,665 --> 00:02:41,865
label. So, you know

67
00:02:42,585 --> 00:02:45,270
and somehow we end up with IP, which

68
00:02:45,270 --> 00:02:46,490
is only two syllables,

69
00:02:47,030 --> 00:02:49,610
and it's shorter to switch based on MPLS,

70
00:02:50,550 --> 00:02:52,629
which is well, I I don't know. That's

71
00:02:52,629 --> 00:02:55,430
an unfortunate anti coincidence or whatever one would

72
00:02:55,430 --> 00:02:56,409
call that, but,

73
00:02:57,510 --> 00:02:58,010
yeah.

74
00:02:58,395 --> 00:03:00,474
Very strange. That's a great that's a great

75
00:03:00,474 --> 00:03:02,235
place to dive in. Like, you know, okay.

76
00:03:02,235 --> 00:03:04,235
What's the problem that was trying to solve?

77
00:03:04,235 --> 00:03:05,055
Right? And,

78
00:03:05,754 --> 00:03:06,414
you know,

79
00:03:06,715 --> 00:03:09,294
look so one one person's perspective,

80
00:03:10,155 --> 00:03:11,534
you know, with incredible

81
00:03:12,074 --> 00:03:13,694
growth of the Internet

82
00:03:14,210 --> 00:03:16,550
in the mid nineties into the late nineties,

83
00:03:17,810 --> 00:03:20,150
we had lots of issues with,

84
00:03:20,450 --> 00:03:23,189
you know, centralized forwarding and software based forwarding.

85
00:03:23,730 --> 00:03:26,032
And just sheer forwarding tables, by the way,

86
00:03:26,032 --> 00:03:28,254
the size of a forwarding forwarding plane. That

87
00:03:28,254 --> 00:03:29,875
size size of forwarding table

88
00:03:30,175 --> 00:03:33,534
and the ability to, you know, have multiple

89
00:03:33,534 --> 00:03:34,435
d s threes

90
00:03:34,735 --> 00:03:37,135
in a single router, you know, or multiple

91
00:03:37,135 --> 00:03:38,115
o c threes.

92
00:03:38,415 --> 00:03:40,034
You know, yes, we're dating ourselves

93
00:03:40,620 --> 00:03:41,180
here. But,

94
00:03:41,740 --> 00:03:45,500
that computational complexity involved with best match what's

95
00:03:45,500 --> 00:03:46,000
that?

96
00:03:46,300 --> 00:03:47,360
Frame relay.

97
00:03:47,659 --> 00:03:49,819
Yeah. Sure. Yeah. Yeah. It's switching. It is

98
00:03:49,819 --> 00:03:50,800
switching. Yeah.

99
00:03:51,580 --> 00:03:53,025
We can go there if you want.

100
00:03:54,064 --> 00:03:55,924
But but the that idea

101
00:03:56,864 --> 00:03:59,504
of forwarding based on fixed length labels that

102
00:03:59,504 --> 00:04:01,044
are always a unique match

103
00:04:01,504 --> 00:04:05,104
versus longest match routing and IP, which has

104
00:04:05,104 --> 00:04:05,604
great

105
00:04:06,610 --> 00:04:07,110
benefits,

106
00:04:07,729 --> 00:04:10,229
you know. However, you know, it's more computationally

107
00:04:10,930 --> 00:04:13,409
complex. You know, the the thought arise that,

108
00:04:13,409 --> 00:04:14,930
you know, if we just took an old

109
00:04:14,930 --> 00:04:15,830
switching paradigm,

110
00:04:16,850 --> 00:04:19,285
and bit forwarded packets based on those fixed

111
00:04:19,285 --> 00:04:21,685
length labels, we can get better throughput in

112
00:04:21,685 --> 00:04:23,285
boxes and keep up with the growth of

113
00:04:23,285 --> 00:04:25,444
the Internet. We also have to remember that

114
00:04:25,444 --> 00:04:27,525
at this time, when MPLS came out, when

115
00:04:27,525 --> 00:04:30,645
Yankov started working on tag switching as we

116
00:04:30,645 --> 00:04:32,824
called it Gotta start there. Yep. Yeah.

117
00:04:35,330 --> 00:04:37,490
ATM, you go back to the Gartner charts,

118
00:04:37,490 --> 00:04:39,970
and ATM was up into the right. We

119
00:04:39,970 --> 00:04:42,069
were gonna have 25 gig

120
00:04:42,610 --> 00:04:44,069
AT or was it Meg?

121
00:04:44,529 --> 00:04:46,210
Maybe it was Meg at the time. It

122
00:04:46,210 --> 00:04:47,430
was Meg at the time.

123
00:04:48,444 --> 00:04:50,225
ATM to the desktop,

124
00:04:50,845 --> 00:04:53,084
and it was going to solve our voice

125
00:04:53,084 --> 00:04:56,285
over IP problems Sure. Because voice can't run

126
00:04:56,285 --> 00:04:57,185
over IP.

127
00:04:57,564 --> 00:05:00,464
It must be on a circuit

128
00:05:00,925 --> 00:05:01,985
based system,

129
00:05:02,444 --> 00:05:05,709
not on a sarcastic packet based system. Like,

130
00:05:05,709 --> 00:05:08,430
you just you can't do this stuff. Right.

131
00:05:08,430 --> 00:05:11,149
And so, you know, MPLS also came a

132
00:05:11,149 --> 00:05:12,829
little bit or a lot out of that

133
00:05:12,829 --> 00:05:13,569
kind of

134
00:05:14,029 --> 00:05:15,329
environmental thinking.

135
00:05:17,225 --> 00:05:18,285
Well, full disclosure,

136
00:05:18,665 --> 00:05:20,985
I I was a four systems trainer, an

137
00:05:20,985 --> 00:05:22,125
ATM switch manufacturer,

138
00:05:23,145 --> 00:05:24,764
before I was a Cisco trainer.

139
00:05:25,305 --> 00:05:28,444
So, I I was talking about NSAP addresses

140
00:05:28,504 --> 00:05:30,365
and RFC fifteen seventy seven,

141
00:05:31,329 --> 00:05:33,110
before I really even learned BGP.

142
00:05:33,410 --> 00:05:33,910
So

143
00:05:35,649 --> 00:05:37,669
Yeah. Okay. That explains

144
00:05:38,529 --> 00:05:39,269
a lot of that.

145
00:05:40,930 --> 00:05:42,229
That's good. So

146
00:05:42,689 --> 00:05:45,095
do you remember the Cisco 8,500?

147
00:05:46,675 --> 00:05:49,335
I The the old 8500?

148
00:05:50,035 --> 00:05:52,355
So does that have a connection to the

149
00:05:52,355 --> 00:05:53,574
Lightstream acquisition?

150
00:05:54,035 --> 00:05:56,134
Because I knew the Lightstream ten ten

151
00:05:56,639 --> 00:05:58,800
and the the ten ten integration into the

152
00:05:58,800 --> 00:06:00,100
Cat 5,500.

153
00:06:00,639 --> 00:06:02,420
Yes. And and I deployed a bunch

154
00:06:02,800 --> 00:06:04,100
of 55 hundreds

155
00:06:04,800 --> 00:06:07,040
at Bell Atlantic way back. So take so

156
00:06:07,040 --> 00:06:10,055
take an take a Lightstream ATM switch. Mhmm.

157
00:06:10,055 --> 00:06:11,274
Put an RP

158
00:06:12,055 --> 00:06:14,634
in the box from a Cisco 7,500

159
00:06:16,055 --> 00:06:17,514
and call it a router.

160
00:06:17,814 --> 00:06:18,314
Sure.

161
00:06:19,095 --> 00:06:19,595
Yep.

162
00:06:20,615 --> 00:06:22,454
Like, man, did that box have a lot

163
00:06:22,454 --> 00:06:23,115
of limitations?

164
00:06:23,654 --> 00:06:25,115
We actually had to invent

165
00:06:26,020 --> 00:06:28,660
games for the MAC address because it didn't

166
00:06:28,660 --> 00:06:30,660
have a large enough MAC address table to

167
00:06:30,660 --> 00:06:31,160
support

168
00:06:31,620 --> 00:06:34,740
in host or larger devices being attached to

169
00:06:34,740 --> 00:06:35,480
the ports.

170
00:06:35,939 --> 00:06:36,439
Like,

171
00:06:36,740 --> 00:06:38,580
we had this whole thing called the adopt

172
00:06:38,580 --> 00:06:39,560
a Mac hack

173
00:06:39,939 --> 00:06:41,925
that we've patented to get to, like, get

174
00:06:41,925 --> 00:06:44,185
around those problems. But, anyway okay.

175
00:06:45,925 --> 00:06:47,764
So that's the origin of a t of

176
00:06:47,764 --> 00:06:50,884
MPLS. I almost said ATM there. Well, anyway

177
00:06:51,605 --> 00:06:53,785
Well, there's there's another another fun,

178
00:06:54,889 --> 00:06:56,889
you know, bash there. You know, what is

179
00:06:56,889 --> 00:06:57,389
MPLS?

180
00:06:58,009 --> 00:07:01,050
It's bringing the complexity of ATM circuit switching

181
00:07:01,050 --> 00:07:05,629
to IP. Yeah. So Yeah. Exactly. Yeah. And

182
00:07:06,009 --> 00:07:08,589
so so those are all the things. Now

183
00:07:08,794 --> 00:07:10,235
if we were to talk about the good

184
00:07:10,235 --> 00:07:12,394
side of MPLS though, first of all, is

185
00:07:12,394 --> 00:07:13,935
MPLS still widely used?

186
00:07:16,394 --> 00:07:16,894
Yes.

187
00:07:17,834 --> 00:07:19,754
Things don't get turned off very quickly in

188
00:07:19,754 --> 00:07:21,834
in most networks, especially here at That was

189
00:07:21,834 --> 00:07:23,779
a kind way to put it, Scott. Thank

190
00:07:23,779 --> 00:07:24,279
you.

191
00:07:24,819 --> 00:07:26,500
Yeah. No. It really is. A lot of

192
00:07:26,500 --> 00:07:28,660
people still use MPLS, but largely for late

193
00:07:28,740 --> 00:07:29,379
for for,

194
00:07:30,339 --> 00:07:33,400
traffic engineering, traffic management, traffic steering

195
00:07:34,100 --> 00:07:36,725
over transit networks, I would say, is the

196
00:07:36,725 --> 00:07:38,665
primary place that I see

197
00:07:39,125 --> 00:07:40,904
MPLS engaged today.

198
00:07:41,285 --> 00:07:43,365
And it's still used for l two VPN,

199
00:07:43,365 --> 00:07:44,745
l three VPN services

200
00:07:45,125 --> 00:07:45,865
no matter

201
00:07:46,165 --> 00:07:47,384
how many times

202
00:07:48,084 --> 00:07:50,345
salesmen say MPLS is dead.

203
00:07:50,750 --> 00:07:51,250
No.

204
00:07:51,629 --> 00:07:54,350
Yeah. MPLS based services are not dead. They're

205
00:07:54,350 --> 00:07:56,529
still out there, and they're still very common.

206
00:07:58,029 --> 00:08:00,430
Well, I think there's an an interesting point

207
00:08:00,430 --> 00:08:02,189
to pick out there on on both of

208
00:08:02,189 --> 00:08:02,930
those applications.

209
00:08:03,870 --> 00:08:05,250
So we started out

210
00:08:06,285 --> 00:08:08,605
trying to keep up with the growth the

211
00:08:08,605 --> 00:08:11,264
bandwidth growth of the Internet core, basically.

212
00:08:12,444 --> 00:08:14,705
And you can correct me if I'm wrong,

213
00:08:15,004 --> 00:08:17,185
but I don't think MPLS VPNs

214
00:08:17,725 --> 00:08:18,865
or traffic engineering

215
00:08:19,660 --> 00:08:23,019
were necessarily those initial goals. And there are

216
00:08:23,019 --> 00:08:25,360
application areas that emerge after,

217
00:08:25,819 --> 00:08:27,980
you know, hey, we found I've got this

218
00:08:27,980 --> 00:08:30,060
label that can be a shim and let

219
00:08:30,060 --> 00:08:32,000
me replicate address space in multiple,

220
00:08:32,539 --> 00:08:34,000
you know, VPNs.

221
00:08:34,379 --> 00:08:36,204
So that was was another one area that

222
00:08:36,204 --> 00:08:37,745
emerged. And then the ability

223
00:08:38,684 --> 00:08:39,664
to very granularly,

224
00:08:41,084 --> 00:08:42,704
like, through through my network,

225
00:08:43,245 --> 00:08:46,284
get better utilization out of underutilized links, you

226
00:08:46,284 --> 00:08:47,804
know, do it very strictly, do it very

227
00:08:47,804 --> 00:08:48,304
loosely.

228
00:08:49,644 --> 00:08:50,544
That's that

229
00:08:50,970 --> 00:08:52,190
that is a secondary

230
00:08:52,570 --> 00:08:55,389
area that also emerged. That's that's my perspective.

231
00:08:56,089 --> 00:08:58,169
Yeah. Yeah. It did not start out. It

232
00:08:58,169 --> 00:08:59,470
started out attacking

233
00:09:00,250 --> 00:09:01,309
the same problem,

234
00:09:02,570 --> 00:09:05,470
that Dino was trying to attack with Lisp,

235
00:09:06,329 --> 00:09:09,625
which which was just sheer the amount of

236
00:09:09,625 --> 00:09:10,445
memory required

237
00:09:11,865 --> 00:09:13,165
on a forwarding device

238
00:09:13,545 --> 00:09:15,004
and the amount of complexity

239
00:09:15,785 --> 00:09:16,445
in building,

240
00:09:18,345 --> 00:09:20,845
a chip that could forward based on

241
00:09:21,990 --> 00:09:22,889
variable length.

242
00:09:23,909 --> 00:09:26,309
That's that's really I mean, and honestly, people

243
00:09:26,309 --> 00:09:28,629
say, why did you have classful? You know,

244
00:09:28,629 --> 00:09:30,549
why did we have classful forwarding in the

245
00:09:30,549 --> 00:09:33,289
early Internet? It's exactly the same answer.

246
00:09:33,669 --> 00:09:36,149
In the early Internet, everything was done in

247
00:09:36,149 --> 00:09:36,649
software.

248
00:09:37,235 --> 00:09:40,035
And the way you make the thing run

249
00:09:40,035 --> 00:09:41,815
faster, forward traffic faster

250
00:09:42,355 --> 00:09:44,915
in software is to be able to say,

251
00:09:44,915 --> 00:09:47,394
oh, that's in this address range. That is

252
00:09:47,394 --> 00:09:48,774
a slash 24.

253
00:09:49,075 --> 00:09:50,995
I don't need a subnet mask. I don't

254
00:09:50,995 --> 00:09:52,590
need to do the XOR operation.

255
00:09:53,210 --> 00:09:55,289
I need to do nothing. I can just

256
00:09:55,289 --> 00:09:57,610
trim stuff off, and I know exactly where

257
00:09:57,610 --> 00:09:59,230
to trim stuff off and ignore

258
00:09:59,610 --> 00:10:01,690
because I know the address range, and I

259
00:10:01,690 --> 00:10:02,669
know it's classful.

260
00:10:02,970 --> 00:10:04,990
So people free, oh, that was so stupid.

261
00:10:05,075 --> 00:10:06,534
No. Actually, at the time,

262
00:10:06,835 --> 00:10:10,054
classful forwarding was really brilliant. It was elegant

263
00:10:10,514 --> 00:10:13,315
and and compact. Yep. Yeah. And it made

264
00:10:13,315 --> 00:10:14,294
forwarding fast.

265
00:10:14,995 --> 00:10:16,995
I I I think it's interesting. We so

266
00:10:16,995 --> 00:10:19,495
in order to make MPLS work,

267
00:10:19,820 --> 00:10:21,419
we had to we had to we had

268
00:10:21,419 --> 00:10:23,899
to virtualize forwarding. We had to create we

269
00:10:23,899 --> 00:10:26,460
had to create an overlay. Right? Because because

270
00:10:26,460 --> 00:10:28,220
of the constraints that we were working inside

271
00:10:28,220 --> 00:10:28,720
of.

272
00:10:29,019 --> 00:10:29,519
And

273
00:10:29,980 --> 00:10:31,980
in in the in the act of creating

274
00:10:31,980 --> 00:10:33,580
an overlay, even at the time it was

275
00:10:33,580 --> 00:10:35,144
being made, I don't think people realize, even

276
00:10:35,144 --> 00:10:36,904
the people building it didn't realize what the

277
00:10:36,904 --> 00:10:39,585
future potential of the technology was. And Sure.

278
00:10:39,705 --> 00:10:41,644
And now there are overlays everywhere.

279
00:10:42,185 --> 00:10:43,245
And, like,

280
00:10:44,105 --> 00:10:46,105
it it took it it took those constraints

281
00:10:46,105 --> 00:10:48,200
and those requirements to kinda dig it out

282
00:10:48,200 --> 00:10:49,980
of us, the this idea of an overlay.

283
00:10:50,200 --> 00:10:53,079
And it's it's hugely important as a as

284
00:10:53,079 --> 00:10:54,679
a concept, as a way of thinking, as

285
00:10:54,679 --> 00:10:55,339
an architecture,

286
00:10:55,799 --> 00:10:58,200
to be able to overlay. Yep. Yeah. Well,

287
00:10:58,200 --> 00:10:59,179
we went from

288
00:11:00,759 --> 00:11:01,980
virtual circuits

289
00:11:02,465 --> 00:11:03,125
in tunnels

290
00:11:03,904 --> 00:11:05,205
to virtual topologies

291
00:11:05,585 --> 00:11:06,404
in overlays.

292
00:11:07,024 --> 00:11:08,485
And that was a huge

293
00:11:08,865 --> 00:11:10,725
mindset shift that MPLS

294
00:11:11,184 --> 00:11:12,725
drove in our heads

295
00:11:13,264 --> 00:11:13,764
about,

296
00:11:14,144 --> 00:11:14,644
no.

297
00:11:15,345 --> 00:11:17,970
Just because it's a tunnel doesn't mean it

298
00:11:17,970 --> 00:11:19,990
has to look like a circuit.

299
00:11:20,690 --> 00:11:23,410
It can be something other than a circuit

300
00:11:23,410 --> 00:11:26,389
and still be a tunnel. I can totally

301
00:11:26,690 --> 00:11:28,149
ignore the source address.

302
00:11:29,195 --> 00:11:32,315
Wait. What? Right. You can do what? And

303
00:11:32,315 --> 00:11:34,315
maybe have no knowledge of it. Yeah. Maybe

304
00:11:34,315 --> 00:11:36,254
even have no knowledge of the source address.

305
00:11:36,315 --> 00:11:37,294
I don't really care.

306
00:11:37,674 --> 00:11:39,134
Is it part of this topology?

307
00:11:39,595 --> 00:11:42,500
Yes. Happy happy joy joy forward away.

308
00:11:42,960 --> 00:11:43,940
I don't care.

309
00:11:44,399 --> 00:11:46,960
Right? And that's that's like a total mind

310
00:11:46,960 --> 00:11:48,000
shift. And by the way, a lot of

311
00:11:48,000 --> 00:11:49,300
people still don't get this.

312
00:11:49,679 --> 00:11:51,059
They still equate tunnels.

313
00:11:51,360 --> 00:11:53,679
That's a really interesting point. I'd never really

314
00:11:53,679 --> 00:11:56,580
connected those dots. Pun completely intended, Tom.

315
00:11:57,345 --> 00:11:57,845
And,

316
00:11:59,345 --> 00:12:01,184
you know, I think there are people that

317
00:12:01,184 --> 00:12:03,584
have other had other problems with other tunneling

318
00:12:03,584 --> 00:12:06,064
protocols and conceptualizing. Like, you know, we've been

319
00:12:06,064 --> 00:12:07,044
doing GRE,

320
00:12:07,824 --> 00:12:08,804
IP and IP,

321
00:12:09,424 --> 00:12:11,664
IPSec, you know, all sorts of other tunneling

322
00:12:11,664 --> 00:12:14,660
mechanisms that it it does take time to

323
00:12:14,660 --> 00:12:16,899
get your head around, you know, and and

324
00:12:16,899 --> 00:12:19,059
that I think you're you're spot on. Right?

325
00:12:19,059 --> 00:12:21,860
The whole like, getting the idea of underlay

326
00:12:21,860 --> 00:12:24,740
versus overlay, that all happened around this time

327
00:12:24,740 --> 00:12:27,160
as well. Yeah. It was driven by MPLS.

328
00:12:27,735 --> 00:12:30,095
Absolutely driven by MPLS. So that's

329
00:12:30,535 --> 00:12:32,455
even if you don't like the technology, even

330
00:12:32,455 --> 00:12:33,975
if you're sitting out there listening to this

331
00:12:33,975 --> 00:12:35,355
episode of the hedge and going,

332
00:12:35,654 --> 00:12:37,274
why do I care about MPLS?

333
00:12:38,054 --> 00:12:39,595
You should at least care

334
00:12:40,259 --> 00:12:42,279
that it created a mind shift

335
00:12:42,980 --> 00:12:45,460
in the way we see networks and the

336
00:12:45,460 --> 00:12:47,559
way we build networks. It made

337
00:12:48,019 --> 00:12:48,679
the entire

338
00:12:48,980 --> 00:12:49,480
concept

339
00:12:50,100 --> 00:12:52,200
of virtual a vertical modularization

340
00:12:52,899 --> 00:12:55,154
possible in our heads. Yep.

341
00:12:56,414 --> 00:12:58,335
And 10 points to your house. That's that's

342
00:12:58,335 --> 00:13:00,654
an that's an important point. Very good. Yeah.

343
00:13:00,654 --> 00:13:01,154
Yeah.

344
00:13:01,615 --> 00:13:03,875
Good. So what else is there, Scott?

345
00:13:04,414 --> 00:13:05,934
I've covered all the good things I can

346
00:13:05,934 --> 00:13:06,509
think of. No.

347
00:13:07,950 --> 00:13:09,950
Well, we, we have more to talk about.

348
00:13:09,950 --> 00:13:12,370
I I I wonder what you you think,

349
00:13:12,429 --> 00:13:13,809
both of you, on

350
00:13:14,269 --> 00:13:15,889
I think around this time,

351
00:13:16,590 --> 00:13:18,910
you know, the whole idea of separation of

352
00:13:18,910 --> 00:13:20,610
a control plane and forwarding plane

353
00:13:21,095 --> 00:13:21,754
was also

354
00:13:22,095 --> 00:13:22,595
introduced,

355
00:13:22,934 --> 00:13:24,875
which I think is at least partly motivated

356
00:13:25,014 --> 00:13:27,434
by chipsets that could handle MPLS.

357
00:13:28,134 --> 00:13:29,274
I, you know,

358
00:13:29,654 --> 00:13:31,014
I spent a lot of time at Juniper,

359
00:13:31,014 --> 00:13:32,695
so I have a very Juniper centric view

360
00:13:32,695 --> 00:13:34,615
of this. I know Juniper did not have

361
00:13:34,615 --> 00:13:36,450
a corner on the market of, you know,

362
00:13:36,450 --> 00:13:38,769
separation of control plane and forwarding plane. I

363
00:13:38,769 --> 00:13:40,230
think they did some other

364
00:13:40,690 --> 00:13:41,190
things

365
00:13:41,889 --> 00:13:43,670
from a router architecture perspective

366
00:13:44,129 --> 00:13:45,590
that let the forwarding plane

367
00:13:46,529 --> 00:13:47,029
forward

368
00:13:47,570 --> 00:13:49,490
and the control plane, you know, handle all

369
00:13:49,490 --> 00:13:52,615
the brainy stuff and then only send those

370
00:13:52,615 --> 00:13:54,535
updates that are needed down to the forwarding

371
00:13:54,535 --> 00:13:57,434
plane when routing topology changes actually

372
00:13:58,134 --> 00:13:59,434
actually motivate that.

373
00:13:59,975 --> 00:14:00,475
Yeah.

374
00:14:00,934 --> 00:14:03,495
SDN started out really as a as an

375
00:14:03,495 --> 00:14:06,019
MPLS technology or as an MPLS as as

376
00:14:06,019 --> 00:14:07,080
we can use MPLS.

377
00:14:07,940 --> 00:14:09,720
And so, yeah, there's a tie there,

378
00:14:10,899 --> 00:14:13,620
between those things. And, again, another tie people

379
00:14:13,620 --> 00:14:15,559
don't seem to recognize or remember,

380
00:14:16,339 --> 00:14:18,740
but that is definitely there was definitely a

381
00:14:18,740 --> 00:14:19,754
tie in that area.

382
00:14:20,315 --> 00:14:22,315
And I think the one of the interesting

383
00:14:22,315 --> 00:14:24,315
lessons too is so the control plane in

384
00:14:24,315 --> 00:14:25,615
MPLS is twofold.

385
00:14:26,154 --> 00:14:29,595
So so there's some sort of carrier, mostly

386
00:14:29,595 --> 00:14:31,455
BGP if we're talking l three VPNs,

387
00:14:31,995 --> 00:14:34,450
and then there's, then there's the actual label

388
00:14:34,450 --> 00:14:36,470
assignment mechanism, mostly LDP,

389
00:14:37,169 --> 00:14:39,409
in the in the case of fairly classical

390
00:14:39,409 --> 00:14:40,230
l three VPNs.

391
00:14:40,610 --> 00:14:42,049
Like and I know there are variations on

392
00:14:42,049 --> 00:14:43,889
this but the the whole idea that you

393
00:14:43,889 --> 00:14:45,029
have two different,

394
00:14:45,615 --> 00:14:47,554
two different algorithms running independently,

395
00:14:48,495 --> 00:14:51,054
doing different things but cooperating to the same

396
00:14:51,054 --> 00:14:51,554
end.

397
00:14:52,014 --> 00:14:53,695
You know, that's a that's a type and

398
00:14:53,695 --> 00:14:56,495
shadow of the orchestration of today and and

399
00:14:56,495 --> 00:14:59,759
it's architecturally correct to let different components do

400
00:14:59,759 --> 00:15:01,759
different things and cooperate with each other that

401
00:15:01,759 --> 00:15:03,200
it doesn't have to be a, you know,

402
00:15:03,200 --> 00:15:03,779
a monolith.

403
00:15:04,160 --> 00:15:05,519
And I think that's I mean, I think

404
00:15:05,519 --> 00:15:07,919
that informs a lot of obviously, it informed

405
00:15:07,919 --> 00:15:10,160
router architecture because if you have these two

406
00:15:10,160 --> 00:15:12,434
things running as separate daemons that that's gonna

407
00:15:12,434 --> 00:15:15,095
drive different, CPU choices and things like that.

408
00:15:15,235 --> 00:15:15,554
But,

409
00:15:16,434 --> 00:15:18,274
that's also I I feel like a like

410
00:15:18,274 --> 00:15:19,495
a fundamental concept.

411
00:15:20,595 --> 00:15:22,455
Well, to to that point, like,

412
00:15:22,835 --> 00:15:23,975
I remember this

413
00:15:24,274 --> 00:15:26,215
moment from twenty five years ago

414
00:15:27,075 --> 00:15:27,575
that

415
00:15:27,970 --> 00:15:30,210
you saw the convergence of how do I

416
00:15:30,210 --> 00:15:30,710
get

417
00:15:31,090 --> 00:15:31,590
RSVP

418
00:15:32,210 --> 00:15:33,909
and BGP to work together?

419
00:15:34,450 --> 00:15:35,269
And when

420
00:15:35,970 --> 00:15:38,769
you know, so Junos by default, you know,

421
00:15:38,769 --> 00:15:41,090
if the BGP next hop is the egress

422
00:15:41,090 --> 00:15:42,070
point of the LSP,

423
00:15:44,035 --> 00:15:45,815
Junos would automatically install

424
00:15:46,355 --> 00:15:48,195
an LSP as the next hop to that

425
00:15:48,195 --> 00:15:50,375
BGP endpoint. And I was like,

426
00:15:51,075 --> 00:15:53,654
woah. It's magic. It just works automatically.

427
00:15:54,274 --> 00:15:57,049
But, like, those operational things matter, right, to

428
00:15:57,049 --> 00:15:59,289
bring the things together to people who don't

429
00:15:59,289 --> 00:15:59,789
understand,

430
00:16:00,409 --> 00:16:02,250
you know, or or have no desire and

431
00:16:02,250 --> 00:16:04,669
need to understand all the protocol mechanisms underneath.

432
00:16:05,289 --> 00:16:07,289
That makes me remember I was working on

433
00:16:07,289 --> 00:16:09,225
a a pretty big MPLS network, and

434
00:16:09,625 --> 00:16:11,225
we were trying to resolve some issue that

435
00:16:11,225 --> 00:16:13,144
hadn't been fully baked before we started putting

436
00:16:13,144 --> 00:16:14,845
configuration down. And and

437
00:16:15,384 --> 00:16:17,865
somebody somebody was like, well, let's just next

438
00:16:17,865 --> 00:16:19,644
top self it so we can go home.

439
00:16:19,704 --> 00:16:22,365
Like Right. Right. Next top self,

440
00:16:23,149 --> 00:16:25,389
facts come up, everything comes up, and traffic

441
00:16:25,389 --> 00:16:26,690
starts flowing. And we're like,

442
00:16:27,149 --> 00:16:27,649
oh.

443
00:16:28,509 --> 00:16:30,429
It was like, yeah. We we should have

444
00:16:30,429 --> 00:16:31,649
already known that. But

445
00:16:32,750 --> 00:16:34,990
I, I wanna poke at Russ here on

446
00:16:34,990 --> 00:16:36,990
something that I think will will get him

447
00:16:36,990 --> 00:16:37,649
to respond.

448
00:16:39,095 --> 00:16:40,134
You know, for all the

449
00:16:41,174 --> 00:16:41,914
out there,

450
00:16:44,294 --> 00:16:46,154
I love the ability

451
00:16:47,254 --> 00:16:50,475
for, you know, BGP to do label distribution.

452
00:16:51,174 --> 00:16:53,274
It's just another NLRI type.

453
00:16:53,610 --> 00:16:56,649
And while some people will use phrases like

454
00:16:56,649 --> 00:16:58,750
the garbage can of routing protocols,

455
00:17:00,889 --> 00:17:02,509
looking looking at mister White,

456
00:17:04,009 --> 00:17:05,950
I see that as elegance

457
00:17:06,490 --> 00:17:07,309
and utility.

458
00:17:07,934 --> 00:17:10,275
Like, the separation of protocol mechanics,

459
00:17:10,894 --> 00:17:11,555
in BGP

460
00:17:12,335 --> 00:17:14,995
versus what I can advertise with it,

461
00:17:15,455 --> 00:17:17,075
dude, I just think that's super useful.

462
00:17:18,335 --> 00:17:20,335
And I think, like, that's the polar opposite

463
00:17:20,335 --> 00:17:20,995
of the mistake

464
00:17:21,849 --> 00:17:22,429
of OSPF,

465
00:17:23,130 --> 00:17:26,429
version two, right, where it could only forward

466
00:17:26,889 --> 00:17:29,630
IPv4 addresses or only deal with IPv4.

467
00:17:30,889 --> 00:17:32,669
A practical decision at the time,

468
00:17:32,970 --> 00:17:33,210
but,

469
00:17:34,169 --> 00:17:36,815
I'm in the I love BGP camp for

470
00:17:36,815 --> 00:17:39,294
how flexible and accommodating it is. So I

471
00:17:39,294 --> 00:17:40,974
would say two things to that. The first

472
00:17:40,974 --> 00:17:43,054
is I still think BGP is the trash

473
00:17:43,054 --> 00:17:44,734
garbage can of the Internet. We still do

474
00:17:44,734 --> 00:17:46,335
way too much with it. I I it

475
00:17:46,335 --> 00:17:47,934
only took till it only took till twenty

476
00:17:47,934 --> 00:17:49,315
six minutes into the recording.

477
00:17:49,774 --> 00:17:50,274
Yes.

478
00:17:51,070 --> 00:17:52,910
But the other thing I'll say is part

479
00:17:52,910 --> 00:17:53,730
of my issue

480
00:17:54,190 --> 00:17:56,269
with the way we've done BGP, and this

481
00:17:56,269 --> 00:17:58,049
started way, way early on,

482
00:17:58,430 --> 00:18:01,330
is that BGP has two ways

483
00:18:02,109 --> 00:18:02,609
to,

484
00:18:04,244 --> 00:18:05,465
to segment data

485
00:18:06,005 --> 00:18:06,984
or to

486
00:18:07,285 --> 00:18:07,785
multiplex

487
00:18:08,085 --> 00:18:09,065
data streams.

488
00:18:09,845 --> 00:18:11,065
The first way is

489
00:18:12,325 --> 00:18:12,825
BGP

490
00:18:13,285 --> 00:18:15,765
has the ability to set up multiple sessions

491
00:18:15,765 --> 00:18:17,785
to carry different kinds of information.

492
00:18:18,309 --> 00:18:18,809
Mhmm.

493
00:18:20,150 --> 00:18:22,630
The second way is is you can create

494
00:18:22,630 --> 00:18:25,690
these address families that were designed originally

495
00:18:26,070 --> 00:18:26,730
to say,

496
00:18:27,269 --> 00:18:29,769
okay, this is IPv4, this is IPv6,

497
00:18:30,230 --> 00:18:32,315
this is x and s. This

498
00:18:32,775 --> 00:18:33,355
is IPX.

499
00:18:33,654 --> 00:18:36,375
This is Vines. I mean, seriously, I mean,

500
00:18:36,375 --> 00:18:38,214
there was all that stuff in there when

501
00:18:38,375 --> 00:18:41,115
not that they they were never codified, but

502
00:18:41,255 --> 00:18:43,109
that was, like, part of the thinking was

503
00:18:43,190 --> 00:18:45,029
we're gonna have different kind. We we live

504
00:18:45,029 --> 00:18:46,950
in a multi protocol world. We need to

505
00:18:46,950 --> 00:18:49,210
be able to carry different address families.

506
00:18:49,910 --> 00:18:51,289
And what we have done

507
00:18:51,670 --> 00:18:54,009
is every time we want to add information

508
00:18:54,069 --> 00:18:54,809
to BGP,

509
00:18:55,349 --> 00:18:57,529
we stick it in as an address family.

510
00:18:57,865 --> 00:19:00,524
Right. With all sorts of magic garbage

511
00:19:00,825 --> 00:19:02,125
to backfill stuff.

512
00:19:02,585 --> 00:19:03,484
And it's like,

513
00:19:03,865 --> 00:19:04,365
no.

514
00:19:05,065 --> 00:19:06,524
I can actually negotiate

515
00:19:06,984 --> 00:19:07,724
in BGP

516
00:19:08,345 --> 00:19:09,804
that I'm going to carry

517
00:19:10,339 --> 00:19:11,960
IP reachability information

518
00:19:12,259 --> 00:19:13,720
in an address family,

519
00:19:14,099 --> 00:19:16,599
and I am going to carry MPLS labels

520
00:19:16,740 --> 00:19:20,039
or security information or whatever I want

521
00:19:20,580 --> 00:19:23,400
in this other stream, which has a completely

522
00:19:23,619 --> 00:19:24,944
different TLV format

523
00:19:25,505 --> 00:19:26,404
that is designed

524
00:19:26,704 --> 00:19:28,325
for that kind of data.

525
00:19:28,704 --> 00:19:31,825
And for some crazy reason, we decided to

526
00:19:31,825 --> 00:19:32,964
make everything

527
00:19:33,424 --> 00:19:34,804
an IP v four address,

528
00:19:35,585 --> 00:19:38,065
like, somehow, like, treat them all like IP

529
00:19:38,065 --> 00:19:40,384
four. Oh, we know how to use address

530
00:19:40,384 --> 00:19:42,859
families. Let's do address families. Let's call starting

531
00:19:42,859 --> 00:19:44,779
with what you know. Like, that's that's there's

532
00:19:44,779 --> 00:19:46,619
a simple answer to that, Russ. Yeah. Come

533
00:19:46,619 --> 00:19:48,140
on, man. But it but it but it

534
00:19:48,140 --> 00:19:49,119
really broke.

535
00:19:49,660 --> 00:19:51,900
If you go look at BGP implementations and

536
00:19:51,900 --> 00:19:54,240
you look at the the switch statement

537
00:19:54,619 --> 00:19:56,480
Right. That handles address families,

538
00:19:57,055 --> 00:19:58,275
That is an extraordinarily

539
00:19:58,894 --> 00:20:00,355
complex piece of code

540
00:20:00,734 --> 00:20:02,914
that could have been utterly avoided

541
00:20:04,575 --> 00:20:05,715
by just saying,

542
00:20:06,174 --> 00:20:08,974
alright. This is an MPLS label. It is

543
00:20:08,974 --> 00:20:10,355
not an address family.

544
00:20:10,940 --> 00:20:12,640
This is different kind of information.

545
00:20:13,180 --> 00:20:14,400
We're going to negotiate

546
00:20:15,019 --> 00:20:15,839
a secondary

547
00:20:16,460 --> 00:20:19,900
TLV or another TLV. Sure. And then, by

548
00:20:19,900 --> 00:20:21,660
the way, you wouldn't have all this garbage

549
00:20:21,660 --> 00:20:23,440
of, are they sending me

550
00:20:23,900 --> 00:20:25,904
are they sending me MPLS labels?

551
00:20:27,025 --> 00:20:28,945
Are they not? You know? What are they

552
00:20:28,945 --> 00:20:30,244
sending me? No.

553
00:20:30,545 --> 00:20:32,005
Did the session negotiate

554
00:20:32,865 --> 00:20:33,365
MPLS

555
00:20:33,664 --> 00:20:34,164
labels?

556
00:20:34,865 --> 00:20:37,424
Yes or no? It's it's not that hard,

557
00:20:37,424 --> 00:20:40,005
Russ. Right? Okay. What is what is the

558
00:20:40,359 --> 00:20:42,599
NLRI type? And the NLRI type is the

559
00:20:42,599 --> 00:20:44,599
discriminator of, like, what is the payload being

560
00:20:44,599 --> 00:20:46,759
carried? So Yes. I know that. I'm not

561
00:20:46,759 --> 00:20:48,700
backing I'm not backing down on this one.

562
00:20:49,480 --> 00:20:51,240
I know that. But let me So let's

563
00:20:51,240 --> 00:20:53,740
let's let's let's twist this. So

564
00:20:54,595 --> 00:20:57,335
I actually had a couple really interesting conversations

565
00:20:57,554 --> 00:20:58,375
last year

566
00:20:59,394 --> 00:21:00,615
with Vint Cerf,

567
00:21:01,075 --> 00:21:03,154
with Jack Haverty, who was,

568
00:21:03,634 --> 00:21:05,414
really early on in EGP,

569
00:21:06,755 --> 00:21:07,974
and then Lynn Bozak.

570
00:21:09,450 --> 00:21:09,950
And,

571
00:21:10,569 --> 00:21:13,690
like, Lynn talked about, you know, the getting

572
00:21:13,690 --> 00:21:14,190
BGP

573
00:21:15,529 --> 00:21:17,929
to be adopted by as many players as

574
00:21:17,929 --> 00:21:18,429
possible

575
00:21:18,730 --> 00:21:21,049
by router manufacturers that we would not even

576
00:21:21,049 --> 00:21:22,909
recognize the names of anymore, Proteon,

577
00:21:24,169 --> 00:21:26,005
Bay Networks, and so forth.

578
00:21:26,625 --> 00:21:27,924
Stop it. Sorry.

579
00:21:29,505 --> 00:21:31,105
They had to they had to be very

580
00:21:31,105 --> 00:21:32,805
minimalistic in what was required.

581
00:21:33,505 --> 00:21:36,644
So and that was, like, what eased adoption.

582
00:21:36,945 --> 00:21:38,545
And I think that might be kind of

583
00:21:38,545 --> 00:21:40,500
a fundamental contributor to some of the things

584
00:21:40,500 --> 00:21:42,440
that you claim are problems or deficiencies.

585
00:21:44,180 --> 00:21:44,680
But

586
00:21:45,619 --> 00:21:47,480
going forward, like, so

587
00:21:47,940 --> 00:21:49,480
we're pretty stuck on BGP,

588
00:21:50,339 --> 00:21:51,720
like it or not, today.

589
00:21:52,684 --> 00:21:54,304
It's not gonna last forever.

590
00:21:55,644 --> 00:21:57,164
So by the way so by the way,

591
00:21:57,164 --> 00:21:59,325
this Scott, this this whole thing about the

592
00:21:59,325 --> 00:22:02,125
way we format stuff is something that Alvaro

593
00:22:02,125 --> 00:22:03,105
and I thought,

594
00:22:04,204 --> 00:22:05,849
like, ten years ago.

595
00:22:06,410 --> 00:22:09,150
We actually wrote drafts Sure. Around,

596
00:22:09,690 --> 00:22:13,230
please stop carrying this stuff as an NLRI.

597
00:22:14,250 --> 00:22:16,970
Please start. There is this other way. We

598
00:22:16,970 --> 00:22:19,710
can just negotiate a different message type.

599
00:22:20,065 --> 00:22:20,565
And,

600
00:22:21,025 --> 00:22:22,545
like, we don't need to put everything in

601
00:22:22,545 --> 00:22:24,565
the same message type. And the community

602
00:22:25,025 --> 00:22:25,525
explicitly

603
00:22:25,825 --> 00:22:27,445
rejected it because,

604
00:22:27,904 --> 00:22:29,045
well, what's the difference?

605
00:22:29,424 --> 00:22:31,045
You know, we can carry

606
00:22:31,744 --> 00:22:33,869
we can carry fake I p v four

607
00:22:33,869 --> 00:22:36,350
addresses with communities if we just wanna carry

608
00:22:36,350 --> 00:22:36,850
communities.

609
00:22:37,309 --> 00:22:39,809
Right. And I'm like, no. Just

610
00:22:40,750 --> 00:22:41,250
stop.

611
00:22:43,390 --> 00:22:46,190
This is the difference between someone who is

612
00:22:46,190 --> 00:22:48,615
writing and maintaining code and someone who is

613
00:22:48,694 --> 00:22:50,234
writing and maintaining philosophy.

614
00:22:52,454 --> 00:22:54,295
Well, there and and to the the point

615
00:22:54,295 --> 00:22:55,434
where I'm trying to

616
00:22:55,734 --> 00:22:57,494
sorry. I'm I'm not trying to sit in

617
00:22:57,494 --> 00:22:58,855
the driver's seat on this. But,

618
00:22:59,734 --> 00:23:01,755
you know, there comes a time when

619
00:23:03,410 --> 00:23:03,910
incremental

620
00:23:04,690 --> 00:23:05,190
adaptation

621
00:23:06,049 --> 00:23:08,630
isn't gonna be sufficient and something new

622
00:23:09,169 --> 00:23:10,549
Mhmm. Is more useful

623
00:23:10,849 --> 00:23:14,470
and that has solid design principles underneath it.

624
00:23:15,884 --> 00:23:18,605
What what comes next after BGP, and what's

625
00:23:18,605 --> 00:23:20,445
it gonna take? And I don't know. Like,

626
00:23:20,445 --> 00:23:22,464
I haven't I haven't thought about that much,

627
00:23:23,644 --> 00:23:26,384
before five minutes ago. Again, yeah, we tried.

628
00:23:27,244 --> 00:23:29,085
There was an attempt to make a a

629
00:23:29,085 --> 00:23:31,265
follow on version of BGP, but

630
00:23:32,090 --> 00:23:34,809
the community is is not interested in such

631
00:23:34,809 --> 00:23:37,450
things at this point. Maybe in twenty years,

632
00:23:37,450 --> 00:23:40,509
they'll be interested. There were several attempts, actually.

633
00:23:40,730 --> 00:23:42,029
One of which was SOBGP

634
00:23:42,410 --> 00:23:44,190
and past eight vectors, by the way.

635
00:23:44,809 --> 00:23:47,230
There were others that were proposed by college

636
00:23:48,224 --> 00:23:51,105
colleges to say, you know, research departments to

637
00:23:51,105 --> 00:23:53,424
say, we could do better if we did

638
00:23:53,424 --> 00:23:56,404
x. And so far, the community has said,

639
00:23:56,785 --> 00:23:59,445
no thanks. Even if it's incrementally deployable,

640
00:24:00,384 --> 00:24:02,644
which which is, okay. Fine. Whatever.

641
00:24:03,079 --> 00:24:04,919
It is a heavy lift. Like, this this

642
00:24:05,079 --> 00:24:06,920
it it's gonna take there have to be

643
00:24:06,920 --> 00:24:07,980
significant value

644
00:24:08,920 --> 00:24:12,119
in migrating to bgp. Next, you know, or

645
00:24:12,119 --> 00:24:13,259
whatever it's called.

646
00:24:14,359 --> 00:24:15,740
May maybe it's gonna,

647
00:24:16,815 --> 00:24:17,454
you know

648
00:24:17,934 --> 00:24:19,794
again, thirty two minutes into the episode,

649
00:24:20,335 --> 00:24:22,174
you know, maybe AI is gonna figure out

650
00:24:22,174 --> 00:24:23,474
a better way to do this.

651
00:24:24,015 --> 00:24:25,934
And once we Oh, no. No. That's scary.

652
00:24:25,934 --> 00:24:27,234
No. Please. Yeah?

653
00:24:28,490 --> 00:24:28,990
Well,

654
00:24:30,809 --> 00:24:31,869
I heard a rumor

655
00:24:32,570 --> 00:24:34,269
that the AI was hallucinating

656
00:24:34,730 --> 00:24:35,710
about the reachability

657
00:24:36,090 --> 00:24:37,230
of a given destination.

658
00:24:39,210 --> 00:24:41,465
Yeah. Sure. That's called a black hole.

659
00:24:42,265 --> 00:24:44,365
And that's that's not artificial intelligence.

660
00:24:45,305 --> 00:24:46,845
That's real stupidity.

661
00:24:48,745 --> 00:24:51,465
So Gives us things to route routing by

662
00:24:51,465 --> 00:24:51,965
rumor.

663
00:24:53,945 --> 00:24:54,924
Routing by insanity.

664
00:24:55,384 --> 00:24:55,884
So

665
00:24:56,400 --> 00:24:58,160
I I I think, you know, there might

666
00:24:58,160 --> 00:24:59,840
come a day when cars have less or

667
00:24:59,840 --> 00:25:01,680
more than four wheels, but it's not today

668
00:25:01,680 --> 00:25:03,519
and it's not within our lifetime. That's the

669
00:25:03,519 --> 00:25:05,440
sort of where this falls in my opinion.

670
00:25:05,440 --> 00:25:06,259
But Gotcha.

671
00:25:07,680 --> 00:25:09,519
Yeah. Don't worry. I don't it doesn't keep

672
00:25:09,519 --> 00:25:10,420
me up at night.

673
00:25:11,164 --> 00:25:12,924
But there's a there's a really interesting guy

674
00:25:12,924 --> 00:25:15,585
from Arista who is out there talking about,

675
00:25:15,644 --> 00:25:17,505
you know, this this is a possibility.

676
00:25:18,845 --> 00:25:20,784
And and look, I'm all for,

677
00:25:22,204 --> 00:25:24,384
give me the right tools to optimize,

678
00:25:25,325 --> 00:25:26,704
traffic flow in my network.

679
00:25:27,139 --> 00:25:29,299
Right? And I don't care what those tools

680
00:25:29,299 --> 00:25:30,359
are based on computationally

681
00:25:30,980 --> 00:25:31,960
if they're reliable

682
00:25:32,339 --> 00:25:34,339
and they do a better job more quickly

683
00:25:34,339 --> 00:25:35,559
than a human does.

684
00:25:36,259 --> 00:25:38,440
So, like, I'm all for that,

685
00:25:39,139 --> 00:25:40,440
you know, again,

686
00:25:40,884 --> 00:25:42,345
that criteria that I just stipulated.

687
00:25:42,804 --> 00:25:44,984
Yeah. So going back to MPLS.

688
00:25:45,365 --> 00:25:47,204
So we've talked about some of the reason

689
00:25:47,204 --> 00:25:47,944
it was invented.

690
00:25:49,365 --> 00:25:51,605
We we've almost touched on a little bit

691
00:25:51,605 --> 00:25:54,359
about why people why people don't like it,

692
00:25:56,119 --> 00:25:58,599
which rightly or wrongly, there are reasons that

693
00:25:58,599 --> 00:26:01,319
people people consider it more complex. It's a

694
00:26:01,319 --> 00:26:03,720
new forwarding plane. You know, if you look

695
00:26:03,720 --> 00:26:04,220
at

696
00:26:04,759 --> 00:26:06,220
the way we build networks,

697
00:26:07,264 --> 00:26:08,625
and this is part of the problem with

698
00:26:08,625 --> 00:26:10,144
v six as well. This is not an

699
00:26:10,144 --> 00:26:12,085
MPLS alone problem. Sure.

700
00:26:13,664 --> 00:26:14,884
We built networks

701
00:26:15,345 --> 00:26:17,365
the way they were originally designed was

702
00:26:17,744 --> 00:26:18,804
IP abstracts

703
00:26:19,200 --> 00:26:20,019
all transports.

704
00:26:20,559 --> 00:26:22,399
And that's not really that important to us

705
00:26:22,399 --> 00:26:24,419
today because Ethernet's pretty universal.

706
00:26:24,960 --> 00:26:26,980
But back in the day, we had ARCNET,

707
00:26:27,119 --> 00:26:29,299
token bus, token ring. Right.

708
00:26:30,000 --> 00:26:30,500
Aloha,

709
00:26:31,039 --> 00:26:32,865
like, all of this stuff.

710
00:26:33,424 --> 00:26:35,904
Stuff, frame relay. There was, you know, o

711
00:26:35,904 --> 00:26:38,545
c three, d s ones, t ones. Like,

712
00:26:38,545 --> 00:26:40,404
there were all these things, PPP,

713
00:26:41,105 --> 00:26:42,865
all this stuff that you had to transport

714
00:26:42,865 --> 00:26:43,605
on top of.

715
00:26:43,904 --> 00:26:46,785
So having an application developer write to every

716
00:26:46,785 --> 00:26:47,605
one of those

717
00:26:48,039 --> 00:26:50,519
was an unhappy thing. Not it doesn't scale.

718
00:26:50,519 --> 00:26:52,759
Yeah. It doesn't scale. So it's much happier

719
00:26:52,759 --> 00:26:54,920
to have a single wasp waste or a

720
00:26:54,920 --> 00:26:57,559
single point where everything comes to, which is

721
00:26:57,559 --> 00:26:58,059
IP.

722
00:26:58,440 --> 00:26:59,684
Yep. So one of the complexities

723
00:27:06,805 --> 00:27:08,965
WASP waste. You are trying to replace the

724
00:27:08,965 --> 00:27:09,465
thing

725
00:27:09,845 --> 00:27:12,025
that was put in place to to

726
00:27:12,404 --> 00:27:13,740
protect the transports

727
00:27:19,099 --> 00:27:19,599
transports.

728
00:27:19,980 --> 00:27:21,839
And so you're you're, like, in the middle

729
00:27:22,059 --> 00:27:23,819
going, yeah. Well, this is the part that

730
00:27:23,819 --> 00:27:25,179
makes the whole thing simple. It's like the

731
00:27:25,179 --> 00:27:27,684
middle Jenga block. Right? It's like I'm gonna

732
00:27:27,684 --> 00:27:29,365
pull the Jenga block out and put a

733
00:27:29,365 --> 00:27:30,264
new one in.

734
00:27:30,644 --> 00:27:32,904
Yeah. Well, okay. Good luck with that one.

735
00:27:33,044 --> 00:27:35,605
And so that's part of the reason that

736
00:27:35,605 --> 00:27:36,105
MPLS

737
00:27:37,204 --> 00:27:40,349
struggled, I think, early on. And in fact,

738
00:27:40,349 --> 00:27:42,430
it's part of the reason that IP v

739
00:27:42,430 --> 00:27:43,329
six has struggled.

740
00:27:43,630 --> 00:27:45,650
Another reason I think MPLS struggled

741
00:27:45,950 --> 00:27:48,430
was because of the sheer complexity of label

742
00:27:48,430 --> 00:27:48,910
swapping

743
00:27:49,390 --> 00:27:51,809
Mhmm. And the label space not being unique.

744
00:27:52,214 --> 00:27:54,855
Yeah. That is so hard for people to

745
00:27:54,855 --> 00:27:57,515
get their head around. People don't even understand

746
00:27:57,575 --> 00:27:59,035
net addresses in ISIS.

747
00:27:59,494 --> 00:28:00,714
Right. Right. Like,

748
00:28:01,414 --> 00:28:04,214
you're getting with the label? Yeah. I'm good

749
00:28:04,214 --> 00:28:07,230
with NSAP addressing. You know, my ATM background.

750
00:28:07,230 --> 00:28:08,590
We we can go there if you want.

751
00:28:08,590 --> 00:28:10,750
But, like, I don't need I don't need

752
00:28:10,750 --> 00:28:12,990
an end to end label value. Right? Like,

753
00:28:12,990 --> 00:28:14,269
it's you know, I

754
00:28:14,830 --> 00:28:16,529
the operation on the router

755
00:28:16,990 --> 00:28:19,009
is, you know, read it, pop it,

756
00:28:19,335 --> 00:28:19,734
you know,

757
00:28:20,375 --> 00:28:21,974
push the new label on and send it

758
00:28:21,974 --> 00:28:24,534
on its way. And, of course, you know,

759
00:28:24,534 --> 00:28:26,615
the real value in MPLS is being able

760
00:28:26,615 --> 00:28:28,875
to learn terms like penultimate hot popping.

761
00:28:30,375 --> 00:28:32,134
That's just fun to say. So

762
00:28:33,319 --> 00:28:35,480
Yeah. But I do substantively, let me talk

763
00:28:35,480 --> 00:28:37,500
about the two things you just brought up.

764
00:28:38,759 --> 00:28:39,740
Was is MPLS

765
00:28:40,279 --> 00:28:42,380
does it add complexity on top of IP?

766
00:28:42,599 --> 00:28:43,099
Absolutely.

767
00:28:43,559 --> 00:28:44,059
Right?

768
00:28:45,400 --> 00:28:47,819
But there is a generational thing on,

769
00:28:48,325 --> 00:28:50,744
you know, okay, who's it complex for?

770
00:28:51,365 --> 00:28:53,684
Right? And the older guys with the white

771
00:28:53,684 --> 00:28:56,345
stuff in their beard and multiple grandchildren,

772
00:28:57,845 --> 00:28:58,505
you know,

773
00:28:59,205 --> 00:29:01,125
that was new stuff that we needed to

774
00:29:01,125 --> 00:29:03,465
learn, figure out how this glues together.

775
00:29:04,490 --> 00:29:07,129
Whereas people who came into the industry after

776
00:29:07,129 --> 00:29:09,869
MPLS had is, you know, more widely adopted,

777
00:29:11,210 --> 00:29:13,609
probably had less of a difficult time getting

778
00:29:13,609 --> 00:29:16,169
their heads around it. I'm gonna I'm gonna

779
00:29:16,169 --> 00:29:17,869
stick by that statement. So,

780
00:29:18,914 --> 00:29:20,215
but let me also say

781
00:29:20,755 --> 00:29:22,595
and and I'm gonna get in trouble with

782
00:29:22,595 --> 00:29:24,615
some of the v six folks listening.

783
00:29:27,394 --> 00:29:30,934
I should only be willing to increase complexity

784
00:29:31,075 --> 00:29:32,455
when it adds value.

785
00:29:33,154 --> 00:29:34,250
Yep. Absolutely.

786
00:29:34,710 --> 00:29:36,250
Absolutely correct. Did MPLS

787
00:29:36,950 --> 00:29:37,849
add value?

788
00:29:39,509 --> 00:29:40,009
Yeah.

789
00:29:41,029 --> 00:29:42,789
We also have to discriminate, okay, what were

790
00:29:42,789 --> 00:29:45,049
really hardware advancements versus MPLS.

791
00:29:45,670 --> 00:29:47,430
But I think you make a pretty clear

792
00:29:47,430 --> 00:29:48,410
argument that

793
00:29:49,045 --> 00:29:51,144
there was value added

794
00:29:51,445 --> 00:29:53,765
and why it got adoption in carrier networks

795
00:29:53,765 --> 00:29:54,825
and core networks

796
00:29:55,285 --> 00:29:56,025
pretty quickly.

797
00:29:58,325 --> 00:29:59,065
V six

798
00:29:59,445 --> 00:30:00,585
is more complex,

799
00:30:02,069 --> 00:30:04,169
and the value has not been

800
00:30:04,549 --> 00:30:05,049
clear,

801
00:30:05,829 --> 00:30:07,589
at least not in the early days. I

802
00:30:07,589 --> 00:30:09,909
mean, when you get to address starvation, you

803
00:30:09,909 --> 00:30:11,269
know, for v four, then you have a

804
00:30:11,269 --> 00:30:12,970
real, you know, value add.

805
00:30:13,429 --> 00:30:14,890
But the first twenty years

806
00:30:15,214 --> 00:30:16,734
were like, I don't see the value of

807
00:30:16,734 --> 00:30:19,214
of adding this complexity. Well, there's there's more

808
00:30:19,214 --> 00:30:21,634
complexity than value in many cases.

809
00:30:21,934 --> 00:30:23,555
And that's partially because

810
00:30:24,255 --> 00:30:24,914
the community

811
00:30:25,295 --> 00:30:25,795
decided,

812
00:30:26,575 --> 00:30:27,795
for whatever reason,

813
00:30:28,174 --> 00:30:30,414
that they weren't just gonna increase the address

814
00:30:30,414 --> 00:30:30,914
space.

815
00:30:31,289 --> 00:30:34,029
They were going to eliminate network address translation,

816
00:30:34,410 --> 00:30:36,750
and they were going to eliminate fragmentation.

817
00:30:37,369 --> 00:30:38,890
And they were going to do this, and

818
00:30:38,890 --> 00:30:40,730
they were going to do that. And you

819
00:30:40,730 --> 00:30:42,970
end up with this huge long list of

820
00:30:42,970 --> 00:30:44,695
stuff that you're going to do,

821
00:30:45,234 --> 00:30:47,315
and you end up with this much more

822
00:30:47,315 --> 00:30:50,055
complex transport. We're gonna get rid of DHCP.

823
00:30:51,474 --> 00:30:51,974
Why?

824
00:30:52,515 --> 00:30:53,015
Why?

825
00:30:53,634 --> 00:30:56,455
Right. Why does it work? Yes. It works.

826
00:30:57,234 --> 00:30:59,230
Why are you getting rid of it? Sure.

827
00:30:59,230 --> 00:31:01,169
You wanna add auto calculation?

828
00:31:01,710 --> 00:31:03,950
Got it. Perfectly fine. Add it on. Make

829
00:31:03,950 --> 00:31:06,509
it optional. Whatever you wanna do. Right. But

830
00:31:06,509 --> 00:31:08,669
for for ten or fifteen years during the

831
00:31:08,669 --> 00:31:11,309
entire development process, the v six folks were

832
00:31:11,309 --> 00:31:12,849
saying no more DHCP.

833
00:31:13,585 --> 00:31:15,825
We're not even gonna write DHCP for v

834
00:31:15,825 --> 00:31:18,164
six. We don't like it. And, like,

835
00:31:19,025 --> 00:31:21,265
who do you think is adopting that? Right.

836
00:31:21,265 --> 00:31:23,205
Who do you think is gonna deploy that?

837
00:31:23,424 --> 00:31:25,200
Well, this this is where I think the

838
00:31:25,440 --> 00:31:28,000
comparison between v six and MPLS kinda gets

839
00:31:28,000 --> 00:31:30,159
a little bit strained because in v six,

840
00:31:30,159 --> 00:31:31,919
we ended up like, no. We're not gonna

841
00:31:31,919 --> 00:31:33,119
do that. And then, yeah, I guess we

842
00:31:33,119 --> 00:31:34,720
will. Like, in a lot of those things.

843
00:31:34,720 --> 00:31:37,619
And so and that created, like, that

844
00:31:38,319 --> 00:31:41,224
divergence of opinions created, I would say, complexity.

845
00:31:41,365 --> 00:31:42,744
When you have to choose between

846
00:31:43,204 --> 00:31:45,765
Slack and DHCP v six, it's a choice

847
00:31:45,765 --> 00:31:47,924
nobody ever had to make before. Yeah. And

848
00:31:47,924 --> 00:31:50,005
any anytime you force someone into a path

849
00:31:50,005 --> 00:31:51,204
where, like, okay, now you have to choose.

850
00:31:51,204 --> 00:31:52,164
Oh, now you have to choose again. Now

851
00:31:52,164 --> 00:31:53,759
you have to choose again. Like, that is

852
00:31:53,759 --> 00:31:56,559
real complexity. Yeah. To me, like, swapping out

853
00:31:56,559 --> 00:31:59,359
the forwarding paradigm, that's not anything compared to

854
00:31:59,359 --> 00:32:01,279
the complexity of the operators not gonna make

855
00:32:01,279 --> 00:32:03,380
10,000 choices they never had to make before.

856
00:32:03,599 --> 00:32:05,380
And I I don't know. I I think

857
00:32:05,840 --> 00:32:06,315
the

858
00:32:06,875 --> 00:32:10,234
probably conceptually MPLS was hard because of the,

859
00:32:10,234 --> 00:32:13,514
like, the the label swapping idea that you

860
00:32:13,514 --> 00:32:15,994
don't have a globally unique label space like

861
00:32:15,994 --> 00:32:18,154
that really messes with with people and they're

862
00:32:18,154 --> 00:32:19,754
like wait a minute. Overlay not a not

863
00:32:19,754 --> 00:32:21,434
a tunnel. So then I think those are

864
00:32:21,434 --> 00:32:21,839
the hard

865
00:32:22,319 --> 00:32:24,319
things. Right. Right. And you have to you

866
00:32:24,319 --> 00:32:25,440
have to think what you have to think

867
00:32:25,440 --> 00:32:27,200
more like a router to really get your

868
00:32:27,200 --> 00:32:29,359
head around MPLS, whereas you don't have to

869
00:32:29,359 --> 00:32:30,599
think like a router if you're just doing

870
00:32:30,599 --> 00:32:32,640
a pure IP. You can think like a

871
00:32:32,640 --> 00:32:34,480
like a postal service employee or like a

872
00:32:34,480 --> 00:32:36,640
regular human. Sure. And I think that's, you

873
00:32:36,640 --> 00:32:37,539
know and

874
00:32:37,945 --> 00:32:39,625
sometimes people just don't wanna think like routers

875
00:32:39,625 --> 00:32:41,865
and that's okay. Like, I don't like, sometimes

876
00:32:41,865 --> 00:32:44,265
the best operators are ones that don't think

877
00:32:44,265 --> 00:32:47,065
like routers, I think. So I don't know.

878
00:32:47,065 --> 00:32:47,965
I see there's,

879
00:32:48,744 --> 00:32:50,585
I see the the pushback to it, but

880
00:32:50,585 --> 00:32:51,325
but also,

881
00:32:51,829 --> 00:32:54,150
I don't see like, I disagree myself with

882
00:32:54,150 --> 00:32:55,450
the complexity argument.

883
00:32:55,910 --> 00:32:58,390
I don't I don't see complexity in MPLS.

884
00:32:58,390 --> 00:33:00,309
I see more I see a few more,

885
00:33:00,309 --> 00:33:02,490
not that many, a few more moving parts,

886
00:33:02,950 --> 00:33:04,549
but that stuff had to be in there

887
00:33:04,549 --> 00:33:06,755
anyway. It was just hidden from your view,

888
00:33:06,914 --> 00:33:09,174
where it's exposed to your view with MPLS.

889
00:33:09,454 --> 00:33:13,095
Yes. I I appreciate that perspective.

890
00:33:13,794 --> 00:33:15,974
I I do think I would say literally,

891
00:33:16,434 --> 00:33:17,734
it does add complexity,

892
00:33:18,194 --> 00:33:19,255
but not a

893
00:33:19,740 --> 00:33:22,480
not an additional cognitive load that's impossible

894
00:33:23,099 --> 00:33:25,420
or super, super hard to get. Right? Because

895
00:33:25,420 --> 00:33:27,420
I if I figured it out, most network

896
00:33:27,420 --> 00:33:29,279
engineers can figure it out. So

897
00:33:30,940 --> 00:33:32,404
Yeah. I don't know about that, Scott. But

898
00:33:32,724 --> 00:33:33,625
Yeah. Okay.

899
00:33:34,244 --> 00:33:36,325
But there's one other thing that I think

900
00:33:36,325 --> 00:33:36,825
really

901
00:33:37,605 --> 00:33:40,105
caused a lot of angst around MPLS

902
00:33:40,804 --> 00:33:42,585
and ended up driving it primarily

903
00:33:42,884 --> 00:33:44,670
into the provider space.

904
00:33:45,230 --> 00:33:46,130
And that is

905
00:33:46,509 --> 00:33:48,529
just that some vendors decided

906
00:33:49,470 --> 00:33:51,090
to charge extra

907
00:33:51,710 --> 00:33:54,590
for an MPLS license Yeah. That's true. Which

908
00:33:54,590 --> 00:33:56,990
makes no sense to me. It's never made

909
00:33:56,990 --> 00:33:57,825
sense to me.

910
00:33:58,384 --> 00:34:01,424
The hardware is already able to do the

911
00:34:01,424 --> 00:34:02,804
job. You built

912
00:34:03,105 --> 00:34:05,984
it into your ASIC. You built it into

913
00:34:05,984 --> 00:34:07,204
your forwarding plane.

914
00:34:07,585 --> 00:34:08,885
The code exists,

915
00:34:09,425 --> 00:34:11,605
and then you are going to charge customers

916
00:34:11,664 --> 00:34:13,284
to turn the code on.

917
00:34:14,000 --> 00:34:15,760
At the same time, you're trying to get

918
00:34:15,760 --> 00:34:17,619
customers to adopt this technology.

919
00:34:18,319 --> 00:34:18,819
That

920
00:34:19,119 --> 00:34:20,099
seems pathological.

921
00:34:20,960 --> 00:34:22,000
I can see it though.

922
00:34:22,639 --> 00:34:24,579
Just taking the other side of the argument,

923
00:34:24,800 --> 00:34:27,085
if you're if you're thinking about support costs

924
00:34:27,085 --> 00:34:29,324
from the vendor's perspective, not the user, the

925
00:34:29,324 --> 00:34:30,144
vendor's perspective,

926
00:34:30,525 --> 00:34:32,525
and you give all these free knobs to

927
00:34:32,525 --> 00:34:33,025
someone,

928
00:34:33,724 --> 00:34:35,644
what's gonna happen? People are going to turn

929
00:34:35,644 --> 00:34:36,144
them.

930
00:34:36,445 --> 00:34:39,164
And and people who did never never had

931
00:34:39,164 --> 00:34:41,550
to make a conscious choice cost wise about

932
00:34:41,550 --> 00:34:43,390
what do I do I purchase this feature

933
00:34:43,390 --> 00:34:44,829
or not? And they had nobody had to

934
00:34:44,829 --> 00:34:46,989
justify to anybody. Oh, yeah. MPLS is just

935
00:34:46,989 --> 00:34:48,750
free. I can I can see the reasoning

936
00:34:48,750 --> 00:34:50,050
from a from a support,

937
00:34:50,590 --> 00:34:51,969
you know, a technical assistance

938
00:34:52,430 --> 00:34:54,989
standpoint? Like, what's gonna happen to our load

939
00:34:54,989 --> 00:34:57,014
when these clueless customers call and they turned

940
00:34:57,014 --> 00:34:58,534
on label switching and now they're in this,

941
00:34:58,534 --> 00:35:00,694
like, this this weird situation because they didn't

942
00:35:00,694 --> 00:35:01,194
understand,

943
00:35:01,974 --> 00:35:04,394
how much time are we gonna spend intact

944
00:35:04,454 --> 00:35:05,355
trying to solve

945
00:35:05,655 --> 00:35:07,894
problems from people who just they just didn't

946
00:35:07,894 --> 00:35:08,394
know.

947
00:35:09,940 --> 00:35:11,059
I I don't know. I I think that's

948
00:35:11,059 --> 00:35:12,260
a little bit weak, but I but I

949
00:35:12,420 --> 00:35:13,460
that's the first thing that comes to my

950
00:35:13,460 --> 00:35:14,980
mind is a counterargument. I don't what do

951
00:35:14,980 --> 00:35:17,380
you what do you guys think? So with

952
00:35:17,380 --> 00:35:19,640
my with my tech engineer hat on,

953
00:35:20,900 --> 00:35:22,980
I I I think there's something to what

954
00:35:22,980 --> 00:35:24,924
you're saying, Tom. I don't think it's the

955
00:35:24,924 --> 00:35:27,264
primary motivation. I'll come back to that.

956
00:35:27,644 --> 00:35:28,625
I'll say that,

957
00:35:29,804 --> 00:35:32,944
you know, as in in my tech seat,

958
00:35:33,565 --> 00:35:34,385
every quarter,

959
00:35:34,844 --> 00:35:35,824
we had training

960
00:35:36,204 --> 00:35:38,444
on, you know, the next set of features

961
00:35:38,444 --> 00:35:39,585
coming out in Junos.

962
00:35:40,409 --> 00:35:42,269
And, like, it was some of the richest,

963
00:35:42,329 --> 00:35:43,230
deepest training

964
00:35:43,929 --> 00:35:46,730
that I ever had. It was engineering presenting

965
00:35:46,730 --> 00:35:49,550
to TAC and other other groups within Juniper.

966
00:35:50,010 --> 00:35:50,989
And that's just

967
00:35:51,929 --> 00:35:54,329
of course, there's new features every quarter. And

968
00:35:54,329 --> 00:35:54,730
there's a

969
00:35:55,505 --> 00:35:57,585
you know, that's COGS, cost of goods and

970
00:35:57,585 --> 00:35:58,644
services. Right?

971
00:35:59,505 --> 00:36:01,265
So you just had to be up to

972
00:36:01,265 --> 00:36:03,744
speed on, okay, what's coming in the next

973
00:36:03,744 --> 00:36:05,364
release and be ready for it,

974
00:36:05,824 --> 00:36:08,164
and then know the SMEs and the escalation

975
00:36:08,304 --> 00:36:08,804
team,

976
00:36:09,184 --> 00:36:12,010
you know, the the very, very deep MPLS

977
00:36:12,150 --> 00:36:15,050
guy, the very, very deep ISIS guy, etcetera.

978
00:36:15,590 --> 00:36:16,090
So

979
00:36:16,469 --> 00:36:18,570
I I do think the bigger motivation

980
00:36:19,510 --> 00:36:21,909
is the product managers and the marketers trying

981
00:36:21,909 --> 00:36:23,909
to find ways to monetize the value that

982
00:36:23,909 --> 00:36:25,130
they're bringing to the market.

983
00:36:25,755 --> 00:36:28,094
Rightly or wrongly, I'm stating this without judgment.

984
00:36:28,234 --> 00:36:30,154
We may save judgment for later, or maybe

985
00:36:30,154 --> 00:36:31,614
Russ will take care of all of that.

986
00:36:33,434 --> 00:36:36,315
I I do understand the need to, like

987
00:36:36,474 --> 00:36:40,170
okay. We we're doing really important work that's

988
00:36:40,170 --> 00:36:42,349
enabling value in customer networks.

989
00:36:43,449 --> 00:36:45,550
Why don't we find a way

990
00:36:46,010 --> 00:36:48,829
to capture some of that value as well?

991
00:36:49,797 --> 00:36:51,184
I I think there are okay ways to

992
00:36:51,184 --> 00:36:53,025
do that. I think there are stupid ways

993
00:36:53,025 --> 00:36:55,585
to do that. Yeah. And I'll my my

994
00:36:55,585 --> 00:36:57,824
favorite example of the stupid ways to do

995
00:36:57,824 --> 00:36:58,324
that

996
00:36:58,784 --> 00:36:59,844
was in the initial,

997
00:37:00,224 --> 00:37:02,784
j series, you know, small CPU routers from

998
00:37:02,784 --> 00:37:03,284
Juniper.

999
00:37:03,585 --> 00:37:05,390
Yeah. You had a two port t one

1000
00:37:05,390 --> 00:37:05,890
card,

1001
00:37:06,269 --> 00:37:08,030
and you needed to buy a license to

1002
00:37:08,030 --> 00:37:09,890
activate the second t one port.

1003
00:37:10,750 --> 00:37:12,510
Are you kidding me? Yeah.

1004
00:37:12,990 --> 00:37:15,070
That's that's a bad example. I'm sure Russ

1005
00:37:15,070 --> 00:37:17,230
has other examples. No. No. I'm I I

1006
00:37:17,230 --> 00:37:19,214
think MPLS is a bad example of it

1007
00:37:19,214 --> 00:37:20,194
as well. Like,

1008
00:37:20,654 --> 00:37:22,514
it's in the soft it's in the hardware.

1009
00:37:22,654 --> 00:37:25,375
Like, let it go. Now, you know, by

1010
00:37:25,375 --> 00:37:26,835
the way, when I was in tech,

1011
00:37:27,295 --> 00:37:29,635
we often wanted to convince

1012
00:37:30,494 --> 00:37:32,250
the the product guys

1013
00:37:32,730 --> 00:37:35,070
to ship all of our routers in Chinese

1014
00:37:35,130 --> 00:37:35,949
puzzle boxes

1015
00:37:36,809 --> 00:37:37,710
or or

1016
00:37:38,730 --> 00:37:39,230
locking

1017
00:37:40,090 --> 00:37:40,590
or

1018
00:37:41,530 --> 00:37:44,989
or locking nesting dolls, you know, because, like,

1019
00:37:45,050 --> 00:37:46,809
if you ain't smart enough to open the

1020
00:37:46,809 --> 00:37:47,114
box,

1021
00:37:48,074 --> 00:37:50,494
please don't be calling TAC. Okay?

1022
00:37:52,394 --> 00:37:53,775
That's that's nice.

1023
00:37:54,875 --> 00:37:57,434
And that's why TAC engineers don't have a

1024
00:37:57,434 --> 00:37:59,454
direct line to be good product management.

1025
00:38:01,820 --> 00:38:04,460
Right. Number of times, like, we used to

1026
00:38:04,460 --> 00:38:06,000
get phone calls all the time.

1027
00:38:06,300 --> 00:38:08,780
Well, my router came with a disk, a

1028
00:38:08,780 --> 00:38:10,719
three and a half inch floppy disk,

1029
00:38:11,340 --> 00:38:13,304
but there doesn't seem to be a disk

1030
00:38:13,304 --> 00:38:14,744
drive. Do I have to buy the disk

1031
00:38:14,744 --> 00:38:15,644
drive separately?

1032
00:38:16,025 --> 00:38:18,105
Right. And you're like, I I don't even

1033
00:38:18,105 --> 00:38:20,184
know where to start. I don't even know

1034
00:38:20,184 --> 00:38:21,164
where to start.

1035
00:38:25,809 --> 00:38:26,549
I hear

1036
00:38:27,250 --> 00:38:29,089
you. Oh, Scott. I like your example of

1037
00:38:29,089 --> 00:38:30,849
the the t one is actually exactly what

1038
00:38:30,849 --> 00:38:32,849
MPLS is. Like, the hardware is there. It

1039
00:38:32,849 --> 00:38:35,589
can move the bits. Yep. It's, you know,

1040
00:38:36,049 --> 00:38:37,730
r and d has been done and paid

1041
00:38:37,730 --> 00:38:40,675
for. Like, there is there is no, like

1042
00:38:41,215 --> 00:38:43,614
but but then, you know, the argument to

1043
00:38:43,614 --> 00:38:44,755
that could be, well,

1044
00:38:46,494 --> 00:38:48,815
why and this leads to, like, pay as

1045
00:38:48,815 --> 00:38:50,574
you go and leads to all sorts of

1046
00:38:50,574 --> 00:38:53,760
licensing schemes which are meddling. Totally. Totally. But,

1047
00:38:53,760 --> 00:38:56,480
like, we'll do we'll we just won't make

1048
00:38:56,480 --> 00:38:57,840
anybody pay for it if they don't use

1049
00:38:57,840 --> 00:38:59,840
it, and that's better for everybody. Right? It's

1050
00:38:59,840 --> 00:39:01,360
like, no. No. I

1051
00:39:02,160 --> 00:39:04,160
Yeah. Well, we do the same thing right

1052
00:39:04,160 --> 00:39:06,425
with routing protocols. Right? You buy this version

1053
00:39:06,425 --> 00:39:08,204
of iOS and you get EIGRP.

1054
00:39:08,744 --> 00:39:10,505
You get that buy that version to get

1055
00:39:10,505 --> 00:39:12,525
IS to IS. Like, what

1056
00:39:13,144 --> 00:39:14,744
I you know, I don't know. Yeah. Like,

1057
00:39:14,744 --> 00:39:15,644
I don't understand.

1058
00:39:15,945 --> 00:39:17,085
So that was a,

1059
00:39:17,465 --> 00:39:18,684
you know, a a

1060
00:39:18,994 --> 00:39:21,059
again, you know, on the Juniper versus Cisco

1061
00:39:21,059 --> 00:39:23,219
days. Right? Early on, I would say the

1062
00:39:23,219 --> 00:39:24,039
first ten

1063
00:39:24,980 --> 00:39:26,679
twelve, thirteen years at Juniper,

1064
00:39:27,539 --> 00:39:30,099
we were able with a straight face to

1065
00:39:30,099 --> 00:39:31,539
say, look, it's all in there and you

1066
00:39:31,539 --> 00:39:33,480
don't need to license it. It's it's on.

1067
00:39:33,779 --> 00:39:35,585
And we that that was a great competitive

1068
00:39:36,684 --> 00:39:38,465
value Yeah. Against against Cisco.

1069
00:39:39,644 --> 00:39:40,385
Now that

1070
00:39:41,324 --> 00:39:42,864
static, you know, over time.

1071
00:39:44,605 --> 00:39:45,425
Product management

1072
00:39:45,724 --> 00:39:48,204
does find a way to, like, okay. Maybe

1073
00:39:48,204 --> 00:39:49,930
we need you know, engineering will say we

1074
00:39:49,930 --> 00:39:51,690
need different builds because this code's getting too

1075
00:39:51,690 --> 00:39:53,849
bloated, and then I have to license certain

1076
00:39:53,849 --> 00:39:56,410
features. And so and that's how new companies

1077
00:39:56,410 --> 00:39:56,910
arise.

1078
00:39:59,369 --> 00:40:02,349
Yes. Like Arista. Like Arista. Yes. Exactly.

1079
00:40:02,650 --> 00:40:04,784
Yes. It it it it really is.

1080
00:40:06,304 --> 00:40:08,065
I don't know. And and so, by the

1081
00:40:08,065 --> 00:40:09,824
way, this is the difference between the companies.

1082
00:40:09,824 --> 00:40:11,505
When I say this, people like, what do

1083
00:40:11,505 --> 00:40:14,484
you mean? Cisco is a marketing driven company.

1084
00:40:14,625 --> 00:40:16,800
Yeah. It just is. Always has been.

1085
00:40:17,199 --> 00:40:17,699
Juniper

1086
00:40:18,079 --> 00:40:20,500
was always an engineering driven company,

1087
00:40:21,039 --> 00:40:23,519
and it always was. And so Juniper could

1088
00:40:23,519 --> 00:40:25,760
never market themselves out of a wet paper

1089
00:40:25,760 --> 00:40:26,260
bag.

1090
00:40:26,880 --> 00:40:27,619
And Cisco

1091
00:40:28,079 --> 00:40:31,119
had really brilliant engineers, but the engineers were

1092
00:40:31,119 --> 00:40:33,094
stuck in some little corner in in a

1093
00:40:33,094 --> 00:40:33,994
darkened hallway,

1094
00:40:34,295 --> 00:40:36,855
and the marketing people were determining everything that

1095
00:40:36,855 --> 00:40:37,594
should happen.

1096
00:40:37,894 --> 00:40:39,974
And so, you know, it's just different company

1097
00:40:39,974 --> 00:40:40,474
philosophies.

1098
00:40:41,335 --> 00:40:43,255
You know? It's just different different ways of

1099
00:40:43,255 --> 00:40:45,355
looking at the problem of making money.

1100
00:40:45,815 --> 00:40:48,619
And so, yeah, I don't know. It's funny.

1101
00:40:48,619 --> 00:40:51,340
So I've always I've care like, Juniper, I've

1102
00:40:51,340 --> 00:40:53,119
characterized exactly the same way.

1103
00:40:53,500 --> 00:40:55,340
Cisco, I would have said, is a sales

1104
00:40:55,340 --> 00:40:58,140
driven company. A sales marketing. Yeah. And yeah.

1105
00:40:58,140 --> 00:41:00,400
So those are adjacent enough. The Venn diagram

1106
00:41:00,619 --> 00:41:01,760
has a large intersection.

1107
00:41:02,220 --> 00:41:03,764
Yeah. And I think, you know, you look

1108
00:41:03,764 --> 00:41:05,925
at John Chambers and, you know, having been

1109
00:41:05,925 --> 00:41:08,164
a sales guy into the CEO role for

1110
00:41:08,164 --> 00:41:08,824
so long,

1111
00:41:09,364 --> 00:41:11,284
this shouldn't be a big surprise. Right? Yeah.

1112
00:41:11,284 --> 00:41:13,844
No. It shouldn't be. Yeah. And Jennifer was

1113
00:41:13,844 --> 00:41:15,944
led by engineers. Right? I don't

1114
00:41:16,335 --> 00:41:17,789
yeah. But, you know, that's Up to a

1115
00:41:17,789 --> 00:41:19,789
time. Right? And that those things change over

1116
00:41:19,789 --> 00:41:20,849
time. So Yeah.

1117
00:41:22,030 --> 00:41:24,510
And and, like, there are Nokia comparisons here

1118
00:41:24,510 --> 00:41:26,450
too just from, you know, my time there.

1119
00:41:26,670 --> 00:41:28,369
I would also say they were engineering

1120
00:41:28,910 --> 00:41:29,410
driven.

1121
00:41:31,355 --> 00:41:32,954
How do I say this? I'll just I'll

1122
00:41:32,954 --> 00:41:33,775
just be direct.

1123
00:41:35,355 --> 00:41:38,174
I never dealt with a more technical

1124
00:41:38,714 --> 00:41:40,654
and business savvy centers

1125
00:41:42,154 --> 00:41:42,654
where

1126
00:41:43,674 --> 00:41:44,174
I

1127
00:41:46,570 --> 00:41:48,730
I knew that I was not gonna get

1128
00:41:48,730 --> 00:41:50,030
a feature request implemented

1129
00:41:50,809 --> 00:41:52,650
unless I show that I knew what I

1130
00:41:52,650 --> 00:41:55,070
was talking about and could really justify it

1131
00:41:55,449 --> 00:41:57,869
and had aligned to 10,000,000 NPOs.

1132
00:41:58,875 --> 00:42:01,195
That's an arbitrary number. But there was tech

1133
00:42:01,275 --> 00:42:03,215
technical qualification and business qualification.

1134
00:42:03,755 --> 00:42:04,255
And

1135
00:42:05,114 --> 00:42:06,954
I always felt, you know, like, having spent

1136
00:42:06,954 --> 00:42:07,855
time at

1137
00:42:08,315 --> 00:42:09,855
Cisco, Juniper and Nokia,

1138
00:42:10,554 --> 00:42:11,775
that helped Nokia

1139
00:42:12,074 --> 00:42:12,574
deliver

1140
00:42:13,610 --> 00:42:15,070
code on time with quality

1141
00:42:15,449 --> 00:42:17,929
in ways that Juniper and Cisco really struggled

1142
00:42:17,929 --> 00:42:18,429
with.

1143
00:42:18,730 --> 00:42:19,230
Yeah.

1144
00:42:19,610 --> 00:42:20,110
Yeah.

1145
00:42:20,570 --> 00:42:22,670
So, yeah, corporate corporate culture

1146
00:42:23,050 --> 00:42:25,050
in Cisco, if the salespeople say I have

1147
00:42:25,050 --> 00:42:26,250
$250,000,000

1148
00:42:26,250 --> 00:42:28,364
deal on the table, the code gets done.

1149
00:42:28,684 --> 00:42:29,824
Yeah. Regardless.

1150
00:42:30,204 --> 00:42:31,025
And at Juniper,

1151
00:42:31,324 --> 00:42:34,204
if an engineer says a a distinguished engineer

1152
00:42:34,204 --> 00:42:36,045
says I have a brilliant idea, the code

1153
00:42:36,045 --> 00:42:36,784
gets done.

1154
00:42:37,565 --> 00:42:39,105
Right. That's the way it works.

1155
00:42:39,804 --> 00:42:41,885
And so yeah. So it's it's an interesting

1156
00:42:42,125 --> 00:42:43,965
but, anyway, back to MPLS. So those are

1157
00:42:44,125 --> 00:42:46,260
I think those are the biggest arguments against

1158
00:42:46,260 --> 00:42:47,239
it or the complexity.

1159
00:42:48,099 --> 00:42:50,099
To me, that you're pulling the Jenga block.

1160
00:42:50,099 --> 00:42:52,180
You're trying to pull the the central Jenga

1161
00:42:52,180 --> 00:42:54,200
block out, and that just makes it, like,

1162
00:42:54,579 --> 00:42:55,400
that's a mental

1163
00:42:56,180 --> 00:42:57,800
that's a mental thing for people.

1164
00:42:58,260 --> 00:42:58,760
And

1165
00:42:59,315 --> 00:43:01,235
then the cost. I think that, you know,

1166
00:43:01,235 --> 00:43:02,535
people charging more.

1167
00:43:03,474 --> 00:43:05,894
Have we overcome those? Are people now

1168
00:43:06,275 --> 00:43:09,175
beyond all of that and deploying MPLS freely?

1169
00:43:09,394 --> 00:43:11,255
I think the problem is in most enterprise

1170
00:43:11,315 --> 00:43:13,894
network, it's got gotten such a bad rap

1171
00:43:14,319 --> 00:43:16,339
that nobody even looks at it anymore.

1172
00:43:17,199 --> 00:43:17,699
Yeah.

1173
00:43:18,799 --> 00:43:21,299
Except for maybe the enterprises that operate,

1174
00:43:21,599 --> 00:43:23,539
like, their own core network.

1175
00:43:24,000 --> 00:43:24,500
Mhmm.

1176
00:43:24,960 --> 00:43:26,239
But I would take I would take that

1177
00:43:26,239 --> 00:43:27,760
comment and flip it, you know, for the

1178
00:43:27,760 --> 00:43:28,819
enterprise users.

1179
00:43:30,485 --> 00:43:31,925
Maybe I don't even need to care about

1180
00:43:31,925 --> 00:43:32,425
MPLS.

1181
00:43:32,885 --> 00:43:34,425
I don't care, you know,

1182
00:43:35,045 --> 00:43:37,364
Verizon, AT and T, Comcast, Charter, what you're

1183
00:43:37,364 --> 00:43:38,664
doing in the core network.

1184
00:43:39,285 --> 00:43:40,425
Just meet my

1185
00:43:41,125 --> 00:43:42,505
don't care how you do it.

1186
00:43:42,980 --> 00:43:45,299
Don't care how you market it. Right? And

1187
00:43:45,299 --> 00:43:47,400
if SD WAN can do that for me,

1188
00:43:47,859 --> 00:43:49,539
I'm okay with that too as long as

1189
00:43:49,539 --> 00:43:50,920
my requirements are met.

1190
00:43:51,539 --> 00:43:53,859
I think this it's the scope of the

1191
00:43:53,859 --> 00:43:55,940
problem you're trying to solve is really critical

1192
00:43:55,940 --> 00:43:58,199
here because a small scope, almost always

1193
00:43:58,954 --> 00:44:00,335
almost always excludes

1194
00:44:00,795 --> 00:44:02,554
like it well, I should say a small

1195
00:44:02,554 --> 00:44:05,195
scope makes it easy to exclude MPLS if

1196
00:44:05,195 --> 00:44:06,394
that's what you'd like to do as an

1197
00:44:06,394 --> 00:44:06,894
engineer.

1198
00:44:07,355 --> 00:44:09,835
But there are some problems that you you

1199
00:44:09,835 --> 00:44:11,835
can't honestly say, well, I'll just push it

1200
00:44:11,835 --> 00:44:13,530
to this other part of the network. Like,

1201
00:44:13,530 --> 00:44:15,130
the one that I'm thinking of in my

1202
00:44:15,130 --> 00:44:16,110
experience is,

1203
00:44:16,809 --> 00:44:18,750
when when you have to do traffic engineering

1204
00:44:19,050 --> 00:44:21,130
because of because of latency and when you

1205
00:44:21,130 --> 00:44:23,289
when you have a global network, you can't

1206
00:44:23,289 --> 00:44:25,130
you can't just say I'll throw more bandwidth

1207
00:44:25,130 --> 00:44:27,914
at the problem because latency doesn't doesn't respond

1208
00:44:27,974 --> 00:44:30,375
to bandwidth. Right? So, like It's a sad

1209
00:44:30,375 --> 00:44:31,654
light hop. That's, you know

1210
00:44:32,534 --> 00:44:33,674
Right. Right.

1211
00:44:34,614 --> 00:44:36,474
Light hop popping. But, like,

1212
00:44:36,934 --> 00:44:39,355
but if you so if you're doing business

1213
00:44:39,654 --> 00:44:41,734
in Shanghai and you're doing business all over

1214
00:44:41,734 --> 00:44:42,394
the world,

1215
00:44:42,909 --> 00:44:44,590
and you need and you need packets to

1216
00:44:44,590 --> 00:44:45,869
go a certain way and then you need

1217
00:44:45,869 --> 00:44:47,570
them to fail over in a certain way,

1218
00:44:47,710 --> 00:44:49,550
you can't just say, well, I'd I'll just

1219
00:44:49,550 --> 00:44:51,869
throw bandwidth at the problem. Like, that's that's

1220
00:44:51,869 --> 00:44:54,349
not a solution. There's not a like, the

1221
00:44:54,590 --> 00:44:56,030
eventually, you do have to deal with an

1222
00:44:56,030 --> 00:44:58,201
overlay. You do have to deal with label

1223
00:44:58,201 --> 00:45:00,502
switching. And this label switching isn't the only

1224
00:45:00,502 --> 00:45:02,803
solution, but it's like there are there are

1225
00:45:02,803 --> 00:45:05,105
some problems of such scope that you you

1226
00:45:05,105 --> 00:45:07,406
don't get to just arbitrarily shift the complexity

1227
00:45:07,406 --> 00:45:09,579
somewhere else. You have to actually look at

1228
00:45:09,579 --> 00:45:10,400
the architecture,

1229
00:45:11,260 --> 00:45:12,859
from the ground up. And I think that's

1230
00:45:12,859 --> 00:45:15,420
where that's where MPLS shines when you have

1231
00:45:15,579 --> 00:45:17,500
when you've got these problems that are fairly

1232
00:45:17,500 --> 00:45:19,339
intractable with other things and you've run out

1233
00:45:19,339 --> 00:45:20,639
of other options, then,

1234
00:45:20,940 --> 00:45:22,460
it becomes really good. And then I think

1235
00:45:22,460 --> 00:45:23,659
when you work on those networks, at least

1236
00:45:23,659 --> 00:45:25,635
for me, when you work on those networks,

1237
00:45:25,635 --> 00:45:27,015
you really start to appreciate,

1238
00:45:27,554 --> 00:45:28,054
how

1239
00:45:29,394 --> 00:45:30,054
how how,

1240
00:45:30,675 --> 00:45:32,675
elegant all of this is that you can

1241
00:45:32,675 --> 00:45:34,994
actually design a network that spans the entire

1242
00:45:34,994 --> 00:45:37,980
world that meets your requirements for latency and

1243
00:45:37,980 --> 00:45:39,039
and is survivable.

1244
00:45:39,340 --> 00:45:41,440
Like, that's impressive that you can actually do

1245
00:45:41,820 --> 00:45:42,300
that. And,

1246
00:45:42,860 --> 00:45:44,059
and to do that, you just have to

1247
00:45:44,059 --> 00:45:45,840
have the the right tool for the job.

1248
00:45:46,139 --> 00:45:47,519
Yep. Totally agree.

1249
00:45:47,900 --> 00:45:48,400
Yep.

1250
00:45:49,019 --> 00:45:50,780
Yeah. So sometimes you just have to eat

1251
00:45:50,780 --> 00:45:52,994
the complexity or the perceived complexity whether or

1252
00:45:52,994 --> 00:45:55,174
not it's complexity is another question. But,

1253
00:45:56,434 --> 00:45:58,194
just to get a certain job done, hard

1254
00:45:58,194 --> 00:46:00,534
problems require complex solutions.

1255
00:46:00,835 --> 00:46:02,135
Yep. They just do.

1256
00:46:03,554 --> 00:46:05,974
You know, so sometimes complexity is not bad.

1257
00:46:06,034 --> 00:46:08,170
It doesn't mean you shouldn't contain it. Doesn't

1258
00:46:08,170 --> 00:46:10,269
mean you shouldn't separate it from other complexity.

1259
00:46:10,329 --> 00:46:12,030
Doesn't mean you shouldn't manage it.

1260
00:46:12,730 --> 00:46:14,670
Just means it has to exist.

1261
00:46:14,969 --> 00:46:16,409
It just is what it is. It has

1262
00:46:16,409 --> 00:46:17,070
to exist.

1263
00:46:17,449 --> 00:46:18,269
Yep. Yeah.

1264
00:46:19,864 --> 00:46:22,505
Can I throw one other final hopeful comment

1265
00:46:22,505 --> 00:46:23,324
in on that?

1266
00:46:25,144 --> 00:46:25,644
I

1267
00:46:26,585 --> 00:46:27,964
I think, you know, we've

1268
00:46:28,264 --> 00:46:31,065
we've seen so in our own careers and

1269
00:46:31,065 --> 00:46:33,724
so many of our our colleagues and friends

1270
00:46:34,420 --> 00:46:37,000
develop subject matter expertise in certain domains.

1271
00:46:37,460 --> 00:46:39,619
Right? And we got limits. We got limits

1272
00:46:39,619 --> 00:46:41,619
on the amount of complexity that we can

1273
00:46:41,619 --> 00:46:43,619
process. Maybe Russ doesn't have any limits, but

1274
00:46:43,780 --> 00:46:45,699
No. No. No. No. We have limits. Trust

1275
00:46:45,699 --> 00:46:48,019
me. We have limits. And and one of

1276
00:46:48,019 --> 00:46:49,159
the things I am

1277
00:46:49,644 --> 00:46:51,105
curiously hopeful about

1278
00:46:52,204 --> 00:46:54,445
is the ability of power tooling that comes

1279
00:46:54,445 --> 00:46:55,664
from the AI world

1280
00:46:56,285 --> 00:46:58,305
to span technical domains,

1281
00:46:59,005 --> 00:47:01,724
you know, even within networking, like what's happening

1282
00:47:01,724 --> 00:47:03,825
at my optical layer, what's happening with IP,

1283
00:47:05,049 --> 00:47:07,049
maybe even bringing in data from the security

1284
00:47:07,049 --> 00:47:09,929
world, and maybe even bringing application performance data

1285
00:47:09,929 --> 00:47:11,929
in, and take a more holistic view of

1286
00:47:11,929 --> 00:47:12,589
the problem.

1287
00:47:14,170 --> 00:47:14,409
I'm

1288
00:47:15,210 --> 00:47:17,150
you know, it's not magic, and

1289
00:47:17,609 --> 00:47:18,045
I'm

1290
00:47:18,525 --> 00:47:20,364
not saying this the way that means that

1291
00:47:20,364 --> 00:47:21,885
that that implies you should go get another

1292
00:47:21,885 --> 00:47:23,085
job in a year. That's not what I'm

1293
00:47:23,085 --> 00:47:23,585
saying.

1294
00:47:23,885 --> 00:47:25,744
But I think there's gonna be ability

1295
00:47:26,364 --> 00:47:28,864
to span those technology domains

1296
00:47:29,244 --> 00:47:31,484
and come up with more holistic solutions and

1297
00:47:31,484 --> 00:47:31,984
approaches

1298
00:47:32,460 --> 00:47:33,980
than I think we've ever been able to

1299
00:47:33,980 --> 00:47:35,820
do in the past, and that's pretty exciting

1300
00:47:35,820 --> 00:47:36,400
to me.

1301
00:47:37,820 --> 00:47:39,500
Yeah. I I like thinking about it like

1302
00:47:39,500 --> 00:47:40,460
that. I think there's,

1303
00:47:41,260 --> 00:47:43,920
I think human judgment is is really amazing.

1304
00:47:44,139 --> 00:47:47,164
And the thing that sometimes our our judgment

1305
00:47:47,164 --> 00:47:49,164
isn't we're not able to employ it fully

1306
00:47:49,164 --> 00:47:51,985
because we lack domain expertise in some areas.

1307
00:47:52,204 --> 00:47:54,125
And so we we end up in our

1308
00:47:54,125 --> 00:47:55,885
our own, like you're saying, our own our

1309
00:47:55,885 --> 00:47:56,945
own domain expertise.

1310
00:47:57,260 --> 00:48:00,059
But if you could connect domain expertise of

1311
00:48:00,059 --> 00:48:02,219
of multiple people or if you could, you

1312
00:48:02,219 --> 00:48:04,219
know, basically, if you could bring your human

1313
00:48:04,219 --> 00:48:06,400
judgment, which I a AI will never replace.

1314
00:48:06,460 --> 00:48:08,460
Correct. And AI doesn't have the ability to

1315
00:48:08,460 --> 00:48:08,960
supplant.

1316
00:48:09,260 --> 00:48:11,019
Like, if you could bring your human judgment,

1317
00:48:11,019 --> 00:48:12,800
which is such a powerful machine

1318
00:48:13,315 --> 00:48:15,175
closer to subject matter expertise,

1319
00:48:15,474 --> 00:48:16,675
then I think you could do a lot

1320
00:48:16,675 --> 00:48:18,195
of things. Like, I I I a lot

1321
00:48:18,195 --> 00:48:19,474
of the value of AI, I think, could

1322
00:48:19,474 --> 00:48:22,914
be explaining to someone, like explaining to a

1323
00:48:22,914 --> 00:48:25,414
person who has x x domain expertise.

1324
00:48:26,239 --> 00:48:28,639
Here's here's the actual problem you're trying to

1325
00:48:28,639 --> 00:48:30,480
solve. Yep. Now what do you think human?

1326
00:48:30,480 --> 00:48:32,159
You solve it. I'm just a dumb machine

1327
00:48:32,159 --> 00:48:34,159
that guesses the next word in my large

1328
00:48:34,159 --> 00:48:36,880
language model. That's that's really pretty stupid but

1329
00:48:36,880 --> 00:48:39,359
but the human judgment can really, I think,

1330
00:48:39,359 --> 00:48:40,099
be augmented.

1331
00:48:40,934 --> 00:48:42,535
Yeah. And what we're actually talking about here

1332
00:48:42,535 --> 00:48:44,454
is the first corollary to Keith's Law, by

1333
00:48:44,454 --> 00:48:46,235
the way, in case you wanna put

1334
00:48:46,614 --> 00:48:47,355
a a formal

1335
00:48:47,815 --> 00:48:49,655
A label on it? Did he A label

1336
00:48:49,655 --> 00:48:50,155
on

1337
00:48:50,614 --> 00:48:51,114
it.

1338
00:48:51,494 --> 00:48:53,035
That's the best way to end the episode.

1339
00:48:53,094 --> 00:48:53,594
Sorry.

1340
00:48:57,750 --> 00:49:00,230
Which is that you can know you're a

1341
00:49:00,230 --> 00:49:01,750
part of the network or that you're part

1342
00:49:01,750 --> 00:49:03,829
of the system really, really well. And you

1343
00:49:03,829 --> 00:49:05,610
can know the pieces that are adjacent

1344
00:49:05,910 --> 00:49:06,809
fairly well.

1345
00:49:07,590 --> 00:49:09,849
Beyond that, it's all rumor and pop psychology.

1346
00:49:12,164 --> 00:49:14,325
Yep. I like it. Yep. We got limits.

1347
00:49:14,325 --> 00:49:15,785
We got limits. We have limits.

1348
00:49:18,005 --> 00:49:20,724
Okay. Cool. Alright. Anything else you wanna talk

1349
00:49:20,724 --> 00:49:22,619
about with MPLS, Scott, before we

1350
00:49:23,099 --> 00:49:25,340
go away? We don't have all day.

1351
00:49:25,900 --> 00:49:27,920
I I could continue. But,

1352
00:49:28,380 --> 00:49:30,539
Well, we could we could schedule another five

1353
00:49:30,539 --> 00:49:31,039
episodes

1354
00:49:31,340 --> 00:49:32,400
and make Whatever.

1355
00:49:32,780 --> 00:49:34,619
If you really need to fill up slots

1356
00:49:34,619 --> 00:49:36,480
that badly, Russ, I'm happy now.

1357
00:49:38,644 --> 00:49:40,184
Yeah. Look, it's been an interesting

1358
00:49:41,045 --> 00:49:43,545
technologies to watch over a career.

1359
00:49:44,085 --> 00:49:44,405
And,

1360
00:49:45,125 --> 00:49:46,804
you know, for all of us, it's a

1361
00:49:46,804 --> 00:49:49,864
snapshot in time. There will be something next.

1362
00:49:50,085 --> 00:49:51,224
Right? You know,

1363
00:49:51,619 --> 00:49:53,619
whether it's gonna be, you know, whatever supplants

1364
00:49:53,619 --> 00:49:54,119
BGP,

1365
00:49:55,619 --> 00:49:57,559
what's beyond IP even,

1366
00:49:58,660 --> 00:50:01,380
won't speculate. But, I feel pretty fortunate to

1367
00:50:01,380 --> 00:50:02,739
have had the seat at the table that

1368
00:50:02,739 --> 00:50:04,500
I've had and just to see the pieces

1369
00:50:04,500 --> 00:50:05,699
of it. And and to be able to

1370
00:50:05,699 --> 00:50:06,920
hear from, you know,

1371
00:50:07,364 --> 00:50:09,525
Tom and Russ on stuff like this, it's,

1372
00:50:10,085 --> 00:50:11,625
it's a great spot to be in.

1373
00:50:12,405 --> 00:50:14,025
Yeah. I think there's a big

1374
00:50:14,804 --> 00:50:15,304
part

1375
00:50:16,644 --> 00:50:18,344
of being a network engineer.

1376
00:50:18,885 --> 00:50:20,885
Like, when I tell people, when I'm teaching

1377
00:50:20,885 --> 00:50:23,159
CLO fabrics and I say, I actually used

1378
00:50:23,159 --> 00:50:25,079
to work on a stroger fabric, and they're

1379
00:50:25,079 --> 00:50:26,139
like, wow.

1380
00:50:27,960 --> 00:50:29,339
That's, like, crazy

1381
00:50:29,960 --> 00:50:32,139
that you actually worked on a stroger fabric.

1382
00:50:32,359 --> 00:50:33,420
Well, yeah.

1383
00:50:34,039 --> 00:50:36,454
But, like, that's part of understanding it that

1384
00:50:36,454 --> 00:50:37,994
I don't know how we replicate

1385
00:50:38,454 --> 00:50:39,755
in a younger generation,

1386
00:50:40,614 --> 00:50:41,114
unfortunately.

1387
00:50:41,894 --> 00:50:44,075
It's just the the physical experience

1388
00:50:44,855 --> 00:50:47,175
of taking that can of contact cleaner and

1389
00:50:47,175 --> 00:50:48,315
going over to that

1390
00:50:48,855 --> 00:50:51,329
one in, 100 out stepper motor driven

1391
00:50:52,269 --> 00:50:53,250
stroger switch,

1392
00:50:53,949 --> 00:50:56,829
spraying all the contact. Somebody called in and

1393
00:50:56,829 --> 00:50:59,550
said, my connection to the base commander was

1394
00:50:59,550 --> 00:51:01,550
fuzzy today. Do you think you could fix

1395
00:51:01,550 --> 00:51:03,170
it? Or, you know,

1396
00:51:05,945 --> 00:51:08,344
somebody's gotta do it. Yep. Somebody's gotta do

1397
00:51:08,344 --> 00:51:09,945
it. It's like it was, you know, it's

1398
00:51:09,945 --> 00:51:10,445
crazy.

1399
00:51:10,985 --> 00:51:14,045
So anyway alright. Tom, anything else on MPLS?

1400
00:51:14,265 --> 00:51:16,684
Mostly pointless label switching, nipples.

1401
00:51:18,220 --> 00:51:20,380
No. But I'm I'm working on another another

1402
00:51:20,380 --> 00:51:22,460
mostly pointless. It's just gonna be another clever

1403
00:51:22,460 --> 00:51:24,539
thing we can we can use those letters

1404
00:51:24,539 --> 00:51:25,099
for. So

1405
00:51:26,220 --> 00:51:26,720
MPLS,

1406
00:51:27,099 --> 00:51:28,640
the city in which MPLS

1407
00:51:29,855 --> 00:51:32,414
was Born? Born or moved forward or however

1408
00:51:32,414 --> 00:51:34,815
you wanna put it, hummed into existence or

1409
00:51:34,815 --> 00:51:36,434
whatever you wanna say. Yeah.

1410
00:51:37,054 --> 00:51:37,875
Cool. Alright.

1411
00:51:38,175 --> 00:51:39,775
Scott, where can people find you if they

1412
00:51:39,775 --> 00:51:42,269
want to or follow you? Check check out

1413
00:51:42,269 --> 00:51:42,769
LinkedIn.

1414
00:51:43,070 --> 00:51:45,150
There aren't too many Robon's, and there's only

1415
00:51:45,150 --> 00:51:47,150
one Scott Robon that I've ever come across.

1416
00:51:47,789 --> 00:51:49,390
And you can see other stuff we're doing

1417
00:51:49,390 --> 00:51:50,829
at solutional.com.

1418
00:51:50,829 --> 00:51:51,730
Reach out anytime.

1419
00:51:52,269 --> 00:51:54,369
Cool. Tom, say your word.

1420
00:51:58,565 --> 00:51:59,065
LinkedIn.

1421
00:52:00,244 --> 00:52:02,405
Audrey made him wait a long time that

1422
00:52:02,405 --> 00:52:02,905
time.

1423
00:52:05,684 --> 00:52:06,184
Yep.

1424
00:52:07,409 --> 00:52:08,929
Alright. I'm Russ White. You can find me

1425
00:52:08,929 --> 00:52:10,789
here at the hedge, rule11.tech.

1426
00:52:11,250 --> 00:52:13,250
I don't know. I feel sometimes I feel

1427
00:52:13,250 --> 00:52:15,349
like banana man. He's everywhere. He's everywhere.

1428
00:52:16,049 --> 00:52:17,349
We know that.

1429
00:52:18,449 --> 00:52:20,309
We live in an attention driven

1430
00:52:21,094 --> 00:52:23,574
and so your attention is extremely important. We

1431
00:52:23,574 --> 00:52:24,934
thank you for listening all the way to

1432
00:52:24,934 --> 00:52:26,775
the bitter end of this hedge, and we

1433
00:52:26,775 --> 00:52:28,394
will cut catch you next time.