Re: [Idr] IDR WG LC on draft-ietf-idr-ix-bgp-route-server (2/2/2015 - 2/16/2015)

[Randy Bush <[e-mail]>]

tl;dr: read last para

> I'm in trouble! I am taking the maximum dosage after your flowspec
> comment earlier only to now read you are working on a neutral and
> community ran internet exchange point!

my first such was in the early '90s.  i just don't kiss and tell.

> I dare not read any more incase I see that you've deployed MPLS as
> I'll have to be carried out on a stretcher! ;)

we prostitutes have our limits

> Back on topic I think the changes have been proposed before and
> frankly are well overdue - use case for route servers is clear - even
> if you can get some orchestration/config vending solution getting the
> same solution rolled out to 40 plus IX points, sadly, is very
> unlikely. I am no fan of route servers but they are a needed evil.

the other decade, i spent some time and energy trying to get peers at an
exchange i helped build to do bgp autoconfig from shared data.  it was
too complex for the smaller ones.  i have been the poster child for irr
based configs for over 20 years.  it has not gone well.

route servers with front ends such as ixp-mangler seem to have enough
web gooey that folk can use them.  the problem is that non-congruence of
the data and control planes has consequences which have repeatedly
bitten us, from datakit, atm, ...  so i have ranted aginst them for some
time.  arnold asked what might make me able less uncomfortable with
them.  draft-ymbk-idr-rs-bfd-00.txt attempts to answer that.

but this is probably the wrong venue for an ops discussion.  this is
where we see how many bgp tricks we can add to a simple idea to make it
comfortably complex and fill out the contributors' list.

so again, i think draft-ietf-idr-ix-bgp-route-server is good to go,
unless the authors have the guts to cut 2.3.2.2, Advertising Multiple
Paths.  and i think draft-ymbk-idr-rs-bfd-00.txt might even make it
palatable to me.

randy