Re: [Bier] [pim] PIM light (draft-ietf-pim-light) and PORT (RFC6559)

FWIW, if an implementation choice, I would like to see considerations about
when it should be used and when it might not be required (provide
additional benefits).

Thanks!

Alvaro.

From: Bidgoli, Hooman (Nokia - CA/Ottawa) <hooman.bidgoli@nokia.com>
<hooman.bidgoli@nokia.com>
Date: November 9, 2022 at 10:59:11 AM
To: Toerless Eckert <tte@cs.fau.de> <tte@cs.fau.de>, pim@ietf.org
<pim@ietf.org> <pim@ietf.org>
CC: bier@ietf.org <bier@ietf.org> <bier@ietf.org>
Subject:  Re: [pim] [Bier] PIM light (draft-ietf-pim-light) and PORT
(RFC6559)

Hi Toerless
>
> I had a quick read of RFC6559, first read I don't see an issue supporting
> PIM light for this RFC.
>
> That said I am not in favor of the wording that PIM Light "MUST" be over
> TCP. I think this is a implementation choice.
> PIM light should work over both and it should be a implementation choice.
>
> So I think the wording should be PIM light would work with RFC 6559 as
> well and it is agnostic to the layer 4 transport
>
> Thanks
> Hooman
>
> -----Original Message-----
> From: BIER <bier-bounces@ietf.org> On Behalf Of Toerless Eckert
> Sent: Wednesday, November 9, 2022 5:01 AM
> To: pim@ietf.org
> Cc: bier@ietf.org
> Subject: [Bier] PIM light (draft-ietf-pim-light) and PORT (RFC6559)
>
> Repeating here on the list what i said on the mike @IETF115, PIM, also Cc'
> BIER WG as hopefully a beneficiary of this work (draft-ietf-pim-light).
>
> We really had a lot of reliability problems under reconvergence of PIM
> with large amounts of stte which are well applicable as a problem to the
> target use-cases of PIM light, especially with BIER which will allow up to
> support a lot of state much better. Thousands of PIM joins that under
> routing reconvergence events have to be buffered as a huge burst and/or
> vendor specific pacing that reduces convergence performance.
>
> We solved these problems with mLDP and BGP signaling instead of PIM, and
> we also then solved them for PIM via PIM over TCP (RFC6559).
>
> I would really like to see:
>
> a) The authors/WG check if/what if any issues threre would be to
> use PORT with pIM light. I hope/expect none, but if there are,
> lets discuss.
>
> b) Include a requirement that PIM light MUST default to use PORT
> and MAY support datagram PIM encapsulation.
>
> Aka: i really see no reason to continue to use datagram encap with PIM
> light, so the "MAY" is really just for unforeseen cases.
>
> Cheers
> Toerless
>
> _______________________________________________
> BIER mailing list
> BIER@ietf.org
> https://www.ietf.org/mailman/listinfo/bier
>
> _______________________________________________
> pim mailing list
> pim@ietf.org
> https://www.ietf.org/mailman/listinfo/pim
>