[17699] in North American Network Operators' Group
Re: $110,000 for Gated Source Code
daemon@ATHENA.MIT.EDU (Gregory Mirsky)
Tue Jun 9 13:43:00 1998
Date: Tue, 09 Jun 1998 11:23:09 -0400
From: Gregory Mirsky <gmirsky@BayNetworks.COM>
To: "Mr. Dana Hudes" <dhudes@graphnet.com>
Cc: "Alex P. Rudnev" <alex@Relcom.EU.net>, Scott Mace <smace@intt.ORG>,
brad <brad@poofy.tbn.tm>, nanog@merit.edu
Hi,
I would like to clarify that GateD worked on totally different NSFNET
backbone. What used to be NSFNET backbone now could be compared to
'Net edges in terms of traffic, number of routes and complexity of
policies. That doesn't say that GateD wouldn't work in today's backbone
at all. Probably there will be some problems if you run it as-is in the
core.
Many start-ups are working on scaling and shaping GateD as well as
some other off-the-shelf IP stacks. And we shall see who'll successed.
--
Gregory Mirsky
Bay Networks, Inc. Direct: 978-916-3772
600 Technology Park Drive Fax: 978-670-8760
Billerica, MA 01821 E-mail: gmirsky@baynetworks.com
> I agree with Alex. People don't seem to realize that GateD
> ran the NFSNET backbone for years. Now, the key difference is
> that was the Cornell version and now we have Merit enhancements
> and fixes some of which are quite important. The Cornell version
> is free, though I know there are bugs in it those bugs are
> documented.
> Some of the bugs are not *really* bugs -- most notorious to me is
> the issue of OSPF restart (change your BGP policy, HUP GateD and
> watch
> your entire OSPF cloud lurch under LSDB recalculate because one
> router
> restarted OSPF). Merit fixed GateD to avoid this problem by
> redoing the OSPF code. That work was funded by the GateD
> Consortium, which is where your $10k goes.
>
> So take your choice: Cornell Gated for free with some fixes over
> what ran the NSFNET, or Merit GateD with some real improvements
> for pay unless you get a research license to play with it.
>
> Dana Hudes
> Graphnet
