RE: Routescience?

I would like to hear it too. However, in fairness to Routescience,
did anyone on NANOG previously ask Cisco and Juniper to publicly
talk about everything they (sometimes painfully) learned about
how to create resilient IS-IS/OSPF/BGP implementations? And even
if anyone would ask, are Cisco and Juniper likely to respond
(thereby giving a heads up to their competitors)?

Routescience may or may not have something worthwhile. However,
I would respect their perogative to not say much more on the
NANOG mailing list.

I think the presentation at NANOG would be a great idea, especially
if it is a joint presentation with an ISP evaluating the product.

Prabhu

I would like to hear it too. However, in fairness to Routescience,
did anyone on NANOG previously ask Cisco and Juniper to publicly
talk about everything they (sometimes painfully) learned about
how to create resilient IS-IS/OSPF/BGP implementations? And even
if anyone would ask, are Cisco and Juniper likely to respond
(thereby giving a heads up to their competitors)?

To the credit of Cisco and (later) Juniper IP routing folks, they always
were in a very close touch with ISP backbone engineers. Cisco in
particular publishes quite a lot of technical information.

Which explais why their (and not their competitors) boxes are used to run
the Internet.

Routescience may or may not have something worthwhile. However,
I would respect their perogative to not say much more on the
NANOG mailing list.

Of course.

--vadim

Date: Wed, 22 Aug 2001 17:04:37 -0400
From: "Kavi, Prabhu" <prabhu_kavi@tenornetworks.com>

I would like to hear it too. However, in fairness to Routescience,
did anyone on NANOG previously ask Cisco and Juniper to publicly
talk about everything they (sometimes painfully) learned about
how to create resilient IS-IS/OSPF/BGP implementations? And even
if anyone would ask, are Cisco and Juniper likely to respond
(thereby giving a heads up to their competitors)?

I don't think that anyone requested source code, but some things are
fair game. What sort of routes does it inject? Does one run WCCP
to send traffic to the RS box, which then source-routes it out?

I deliberately picked source-routing as an example; many people
filter both SSRR and LSRR. Reduced effectiveness, but not a problem
yet. However, there are enough boneheads who block all ICMP, meaning
that "source route failed" doesn't make it back, that I'm none too
keen on anything that source-routes. I don't want packets timing
out.

See far we've covered:

1. Make the device a border router
2. Snag packets via WCCP and source-route out
3. Inject longer prefixes on a trial-and-error basis
4. I'm too tired to think of other techniques.

Don't get me wrong... there are several approaches to the problem.
However, avoiding scalability, stability, _and_ interoperability
problems is non-trivial.

Routescience may or may not have something worthwhile. However,
I would respect their perogative to not say much more on the
NANOG mailing list.

Agreed to a certain extent. It's their company, their decision,
and private communication (or no communication) to prevent leaking
trade secrets... perfectly valid. I don't think that we've reached
that point, though.

Consider also that it's not difficult for a competitor with tens of
millions in R&D to buy a box, sniff packets, and do whatever else
is permitted by law. (I have no idea if they have a non-compete
clause in a purchase contract, so I might be off base here. Then
again, IANAL, and don't want to speculate on contract law and what
would hold up.) Anything that is readily observed should, IMHO, be
fair game for discussion.

I think the presentation at NANOG would be a great idea, especially
if it is a joint presentation with an ISP evaluating the product.

Agreed.

Eddy