Internet Core Routing - Ethernet

Mailman List Mirror (Read Only)
1

Folks,

Recently there was a string about L3 Switches in the core. I really don't like this string because I belive it reflects NANOG in a very bad light and I'm going to speak up here.

1. Why are you talking about vendors when you should be talking about technology on this list. Just like the charters. I would recommend to NANOG "if you disagree, say nay" that any post to NANOG with a vendor name in it be sent to the vendor for comment before being posted to the list. Just my 2 cents here.

2. Ethernet is the technology. If you don't see it, you are blind. Let's talk about Internet core routing in those terms. Sure, all vendors suck, but some suck WAY more than others. I think we all know the names. I would love to be challenged on my knowledge of ethernet as a network engineer. I know way more than I ever did about token ring. Wake up. How many routes can vendor X support with IBGP Nailed Routes? What is the convergence time for 100,000 routes? Does vendor Z have wire-speed ACL,s, PBR, MPLS? 10GigE? How much does that cost? Does vendor Y have any reference customers? Let me see the video? How do I manage CAM? I think we all know who the strong vendors are. There are a few that will be around for awhile.

Bobby

2

1. Why are you talking about vendors when you should be talking about
technology on this list. Just like the charters. I would recommend to
NANOG "if you disagree, say nay" that any post to NANOG with a vendor name
in it be sent to the vendor for comment before being posted to the list.
Just my 2 cents here.

Operations has alot to do with experience with specific equipment, more so
then it does abstracted conversations about theoretical configurations.

There are abundant protocols that lack implementation. Many people
believe they can instantiate an implementation by writing an RFC, or so
you would get the impression by reading RFCs.

Regarding "be sent to the vendor for comment before being posted", your
question about "Does vendor Z have wire-speed X?" has produced outright
lies for the last 10 years. There is *always* a catch to any such claims
regarding routers.

Take posts to NANOG with a grain of salt. There is a large range of
experience.

BTW, you listed a bunch of questions. Please post a specific one you want
to know about. You might get it answered.

+----------------- H U R R I C A N E - E L E C T R I C -----------------+

3

technology on this list. Just like the charters. I would recommend to
NANOG "if you disagree, say nay" that any post to NANOG with a vendor name
in it be sent to the vendor for comment before being posted to the list.

Hmm... so if somebody posts to the list with the problem, and somebody else saw
that same issue and got a fix from the vendor, they need to send it to the
vendor for comment, or they can say "Oh, you're being bit by bug (can't say
because it would identify the vendor) in a (vendor model you can't say) several
hops upstream from you".

talk about Internet core routing in those terms. Sure, all vendors suck,
but some suck WAY more than others. I think we all know the names. I would

And how did we learn the names? Let's see.. Cisco, Juniper, Proteon, Bay,
add whatever names you want. You probably have your own opinions on what
each piece of gear is good at.

Now, did you actually *buy* and *use* all of that gear yourself? Or did
you let the common knowledge base called NANOG help you out? Remember that
learning from the mistakes and mishaps of others is a lot cheaper than doing
them yourself...

now way more than I ever did about token ring. Wake up. How many routes
can vendor X support with IBGP Nailed Routes? What is the convergence time
for 100,000 routes? Does vendor Z have wire-speed ACL,s, PBR, MPLS? 10GigE?

Which would you trust for any of these numbers - a salesdroid *saying* a
number, or 5 different people on the NANOG list who have all seen a much
lower-performance number out in the field?

How helpful would this list be if vendors had the right to squelch any
adverse publicity?

2. Ethernet is the technology.

Excuse me if I chuckle, having heard THAT before in the last 2 decades or so.

I've learned to not take *anybody* seriously when they say something is "THE"
anything. Structured programming wasn't the end-all, nor was ATM, nor was
Java, nor will XML or Ethernet. Yes, 10G-E will probably see wide deployment.
But I'll make a prediction - there will be something else coming out to
replace it long before it finishes replacing what's out there now.

(For bonus points, compare the level-1 media characteristics of the original
10mbit-over-thickwire with the 10gig-over-optical, and ask yourself if there's
anything in common other than the name. It's amazingly reminiscent of the
saying "I don't know what language number-crunchers will use 20 years from now,
but it will be called Fortran"....)

4

Thus spake <Valdis.Kletnieks@vt.edu>

> 2. Ethernet is the technology.

Excuse me if I chuckle, having heard THAT before in the last 2 decades or so.

I've learned to not take *anybody* seriously when they say something is "THE"
anything. Structured programming wasn't the end-all, nor was ATM, nor was
Java, nor will XML or Ethernet. Yes, 10G-E will probably see wide deployment.
But I'll make a prediction - there will be something else coming out to
replace it long before it finishes replacing what's out there now.

(For bonus points, compare the level-1 media characteristics of the original
10mbit-over-thickwire with the 10gig-over-optical, and ask yourself if there's
anything in common other than the name.

The electrical characteristics of 10BaseT aren't all that similar to 10Base[25]
either. However, all of the 802.3 variants use CSMA/CD for half-duplex
operation, and 802.11's CSMA/CA is reasonably similar. All of the 802.3 and
802.11 variants use the same MAC and LLC layers.

Sure, the framing and modulation has varied over time. GE's undersize-frame
packing was a neat innovation, 10GE's elimination of half-duplex was a bit
overdue, jumbo frames could be neat if they ever get deployed, and 802.1p/q
opened a lot of doors. However, throughout Ethernet's evolution, it's remained
essentially the same beast from the user's perspective, and the L2 operation is
still the same.

Do I get my bonus points? :slight_smile:

S