Re: [Idr] Mirja Kühlewind's No Objection on draft-ietf-idr-ix-bgp-route-server-11: (with COMMENT)

[Robert Raszuk <[e-mail]>]

Hi Marco,

Altering AS_PATH indeed does not happen that often. Let's put that one
aside keeping in mind that practical use cases exists where adding an AS is
needed which to me precludes "MUST".

But I can't agree with statement that IXP RS "should be as transparent as
possible to route propagation" as every IX which assists in routes handling
offers extensive RS to client (and to some level even client to RS)
filtering policies.

Ref:
https://ams-ix.net/technical/specifications-descriptions/ams-ix-route-servers

That proves black on white that "transparency" is gone. Now the question is
if we should RFC something which would be completely unpractical or
unusable ?

Many thx,
R.


On Mon, Jun 20, 2016 at 11:43 PM, Marco Marzetti <[e-mail]> wrote:

> On 2016-06-20 20:24, Robert Raszuk wrote:
>
>> Jeff,
>>
>> I am somewhat late in chiming in on this thread, but I'm also
>>> supportive of
>>> making this a MUST.
>>>
>>
>> ​As discussed before I disagree. Injecting AS(es) is a simplest form
>> of making some paths less preferable then others. ​And per peer
>> policy asking RS send my paths to peers X,Y, Z with + 4 ASes, but to
>> peer A, B, C do not prepend as Niels points out maybe there already
>> today.
>>
>> If you're intending to inject your AS number into the path, then
>>> you're
>>> acting as a standard BGP speaker, not a Route Server.
>>>
>>
>> ​To me the Route Server is a Route Reflector for EBGP sessions. It
>> really does not matter if you are messing with AS_PATH there or not.
>> What actually matters ​is that you do not set next hop to yourself
>> there.
>>
>
> I would like to underline that this draft depicts the behavior of a route
> server in an IXP environment.
> And, for many reasons, an IXP should be as transparent as possible when it
> comes to route propagation.
>
> --
> Marco
>