Re: [Isis-wg] [Technical Errata Reported] RFC5310 (2462)

["Les Ginsberg (ginsberg)" <[e-mail]>]

Tony -

> -----Original Message-----
> From: Tony Li [mailto:[e-mail]]
> Sent: Friday, August 13, 2010 3:23 PM
> To: Les Ginsberg (ginsberg)
> Cc: Bhatia, Manav (Manav); Chris Hopps; [e-mail] list; Ran
> Atkinson; [e-mail]; RFC Errata System;
> [e-mail] Farrel; [e-mail] White
> Subject: Re: [Isis-wg] [Technical Errata Reported] RFC5310 (2462)
> 
> 
> Les,
> 
> > In the context of the discussion regarding
> > draft-ietf-isis-purge-tlv-03.txt, what is the point of
> > clarifying/correcting RFC5310 in this way given that we now need to
> > update this behavior to allow TLVs other than the authentication TLV
> in
> > purged LSPs?
> 
> 
> Do you have a better way?  There needs to be some text in 5310 for how
> to handle purges.  We don't want to add text that is a forward
> reference to every explicit TLV that can appear in purges in the
> future.  The text that I'm proposing simply puts 5310 on the same
> footing as 5304, which we have to address anyways.

I am thinking of this from what I think is a practical point of view.
RFC5304 updates RFC3567 - which is several years old. I would have
expected that folks who implemented RFC5310 would have used the same
purge strategy that was specified in 5304/3567 - despite the fact that
5310 (quite unintentionally I assume) did not include the necessary
explicit language. If they did, then they are already doing what the
errata is intended to clarify.

If they have not followed the RFC5304 procedures in regards to purge,
then telling them they need to introduce RFC5304 behavior seems
pointless if we intend to "immediately" follow this up by telling them
that the RFC5304 behavior needs to be updated to allow for TLVs other
than the authentication TLV in purge LSPs.

??

> 
> 
> > I would also point out that draft-ietf-isis-mi-03.txt also
stipulates
> > that the instance TLV MUST be retained in purged LSPs (see the last
> > paragraph of Section 2.1) - so we now have two cases that require
> TLVs
> > other than the authentication TLV to be included in purge LSPs.
> 
> 
> Excellent, thanks for pointing that out.
> 
> This suggests that we need a more generic mechanism.  It seems to me
> that the correct approach here is to create or extend an IANA registry
> of TLVs that can be found in purges.  This could become another column
> in the "TLV Codepoint Registry" or another independent registry.
> 

An additional column in the existing registry seems like an excellent
idea to me.

   Les

> RFC's that are then discussing purges could reference this registry.
> 
> Thoughts?
> 
> Tony