Re: [spring] [**EXTERNAL**] Re: [Pals] Martini Pseudowires and SR

["Shah, Himanshu" <[e-mail]>]

Agree with Stewart.
T-PEs could be intra or inter-AS and could get unwieldy when interworking with mixture of
legacy and (newer) lightweight T-LDP peers.


Thanks,
Himanshu


From: Pals <[e-mail]> on behalf of Stewart Bryant <[e-mail]>
Date: Monday, May 30, 2022 at 4:10 AM
To: Andrew Alston - IETF <andrew-ietf=[e-mail]>, [e-mail] <[e-mail]>, mpls-chairs <[e-mail]>
Cc: SPRING WG <[e-mail]>
Subject: [**EXTERNAL**] Re: [Pals] [spring] Martini Pseudowires and SR
Including the PALS and MPLS WGs in the discussion.

In the case of PWs, LDP runs directly between the T-PEs to provide the control plane. If it is known that the only use of LDP is to support PW, then a lightweight profile of LDP might be implemented, ignoring unused parts, but this does not necessarily need a standard.

Before you can profile LDP, you have to also profile PWs to determine which subset of the PW system you need to support. The danger here is that you end up going through the PW development cycle again as old requirements re-emerge.

Stewart



Sent from my iPad


On 30 May 2022, at 07:22, Andrew Alston - IETF <andrew-ietf=[e-mail]> wrote:

Hi All,

Sending this email wearing only the hat of a working group participant.

One of the things that our network uses, and is used by so many networks out there, are martini based pseudowires (which for clarity are generally setup using what is described in RFC8077).  In an SR world however, this creates a problem, because typically you don’t want to run LDP in an SR context.  This means that standard martini pseudowires no longer function.  This gets even more complicated when you want to do martini based pseudowires over an IPv6 only network, particularly considering the lack of widespread support for LDP6.

This is also relevant in cases where networks wish to run SR-MPLS in the absence of SRv6 for whatever reason.

So, my question to the working group is this:

Is it worth looking at creating a form of LDP light – both compatible with IPv4 and IPv6 – that simply exists to setup and tear down the service labels for point to point services.  A form of targeted LDP without all the other complexities involved in LDP – that could potentially run at a lower preference than LDP itself (so if LDP is there, use it, if not use this)

Before I start drafting though, I would like to hear from the working group if there are others who feel that this is worth doing and, call this a call for expressions of interest in those who may be willing to work towards something like this.  Happy to take emails on list or off list and see if we can find a solution.

Looking forward to hearing from you all

Thanks

Andrew



_______________________________________________
spring mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/spring [ietf.org]<https://urldefense.com/v3/__https:/www.ietf.org/mailman/listinfo/spring__;!!OSsGDw!PXv79SX-yRuZt5V8svtex957gTcCXa1ez_vgk91k0NUp5BXkHYadcjJJ1C63l-dp07qx1xdY9OEyxkfNpI85$>