Re: [Idr] One week extension to WGLC for draft-ietf-idr-large-community

[Robert Raszuk <[e-mail]>]

Hi,

If AS1 or AS2 can prepend themselves there is no issue.

However the issue with AS3 offering 1-6 is that if AS2 end point is to
prepend at AS3 just by 1 AS2 would need to delete community  set by AS1 and
replace it by offered by AS3 prepend by 2.

As to the last point Jakob made that AS3 would not execute +1 prepend and
+2 prepend communities it is interesting how does AS3 knows who set each of
those and which one to not execute.

Thx,
R.

On Nov 16, 2016 3:26 AM, "heasley" <[e-mail]> wrote:

> Tue, Nov 15, 2016 at 07:09:03PM +0100, Robert Raszuk:
> > And what stops you from having a bit smarter clauses where you regex
> match
> > on exactly one, two or N occurences of a given community string anywhere
> in
> > your received community list following different local PREPEND action ???
> >
> > Expanded matches for communities have been supported for years.
>
> why would AS3 not just offer communities for 1 - 6 prepends?  at some
> point,
> the number of prepends is pointless and AS1 could cause AS2 to prepend or
> prepend itself to achieve the same effect.
>
> > On Nov 15, 2016 11:28 PM, "i3D.net - Martijn Schmidt" <
> > [e-mail]> wrote:
> >
> > > Hi Robert,
> > >
> > > Route-maps don't work that way in today's router software, every
> clause is
> > > evaluated once and then the route-map execution either stops (if no
> > > continue statement is given) or proceeds to the next clause (if a
> continue
> > > statement is given). Either way, the community will only be matched
> once
> > > even if it's been tagged twice on the same prefix.
> > >
> > > Best regards,
> > > Martijn
> > >
> > > On 11/15/2016 03:19 PM, Robert Raszuk wrote:
> > >
> > > Hi Bruno,
> > >
> > > Since I was going to ask the same question let me provide an example:
> > >
> > > AS1 ---- AS2 ---- AS3 ---- peer_X
> > >
> > > AS3 publishes community with action if I receive this community I will
> > > PREPEND my AS# one time towards my peer X.
> > >
> > > Now AS1 comes with a need for such prepend on PI prefix P. Then AS2 for
> > > perhaps completely different reason also comes up with idea to hint
> AS3 to
> > > execute prepend on the same very prefix P.
> > >
> > > So AS3 is receiving the two communities which both are valid and truely
> > > intended. Why we would drop it ?
> > >
> > > I admit when the topic came up originally the major concern to me was
> to
> > > prevent stuffing community with N identical sets of values by same AS.
> > >
> > > But even this case can be legitimate. Let's consider that said AS2
> needs
> > > prepend by 2 to some prefixes and AS3 offers *only* community of single
> > > PREPEND. Why not adding it twice to such selected prefixes not be a
> way to
> > > signal such requirement ?
> > >
> > > Regards,
> > > R.
>