On Mar 26, 2009, at 4:41 PM, Christopher Morrow wrote:
Danny's slides were not posted is there an internet draft
documenting those
BGP "local" optimizations leading to system wide systemic issues ?
I provided them to the chairs, they are available now.
What is the point of this discussion ?
There are many points. Mainly that there is room for
improvements in three places: operations and network
architecture, implementations, and BGP itself.
*A* To change reflector spec back to prohibit reflection of paths
previously
received from the originating client ? Could such recommendation
also take
into the account the impact for the reflector itself ?
Sure.
*B* To kill some other work (ADVERTISE_OWN) which while building on
reflecting back to the originating client due it's design not having
anything to do with plain IPv4/IPv6 routes - hence have no local PE
box nor
system wide impact ?
See email I sent a moment ago. Like I said, I have
zero issue with a solution such as that, so long as
implementers aren't planning to change general BGP
behavior to make it easier.
I'll work with a few folks to document several of these
issues in a draft or two, apologies for not having them
available sooner.
-danny
_______________________________________________
GROW mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/grow
