[quagga-dev 8211] Re: request for testing

Alexis Rosen quagga-users at alexis.users.panix.com
Fri Sep 3 09:56:06 BST 2010


On Sep 1, 2010, at 4:46 AM, paul at jakma.org wrote:
> On Fri, 27 Aug 2010, Denis Ovsienko wrote:
> 
>> This task is best solved not by software, but by a live person, the network administrator, which can contact remote side to find out, if they execute their side of the peering contract properly or not. Let computers do bit-crunching as defined by the technical spec.
> 
> This has been discussed on the IDR list too. The thinking there now is that just the affected path should be "broken" - the session shouldn't be taken down. That is, you shouldn't break /all/ paths just cause one path has bad attrs - particularly not for optional attributes which can 'tunnel' across many routers (which dont recognise them, and can't recognise they're invalid). E.g. something along the lines of:
> 
> https://datatracker.ietf.org/doc/draft-ietf-idr-optional-transitive/

I've been meaning to write about this...

I know that Quagga is being used by a lot of exchanges and peering points, but it's also heavily used by end nodes - people who aren't providing transit to anyone. And for those users, it's a good bet that it's almost *never* the right choice to take down the session, because for them, that translates to "the internet is broken" rather than "I can't reach one site". For example, when we first hit the AS4 crasher (18 months ago? I forget), it took a bunch of sites completely off the net.

It should be at least an option for such sites to prevent malformed routes from taking down an entire session.

/a


More information about the Quagga-dev mailing list