Re: [rbridge] [Isis-wg] Progressing draft-ietf-isis-trill

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

Erik -

My concern in regards to the proposed use of two documents
(draft-ietf-trill-adj/draft-ietf-isis-trill) has nothing to do with
"process" - which you discuss in detail below. It has to do with the
interdependence of the content. When defining extensions to the IS-IS
protocol it is necessary to define the new protocol behaviors and to
define the new TLVs, sub-TLVs, etc. that are used in support of the new
behaviors. To separate the two into different documents is arbitrary and
illogical. Until the review of the protocol behavior is completed we
cannot know if the current definition of the TLVs is complete and
accurate. And until the review of the TLVs is completed we cannot know
whether the protocol behavior description is complete and accurate. Thus
the separation makes life more difficult for both the reviewers and the
authors - and ultimately makes the use of the document(s) as a reference
more difficult.

I used the example of RFC 5303 - but one could look at any of the RFCs
that have been produced by the IS-IS WG (as well as other WGs) over the
years. No one has chosen to make such a separation - and I think for
very good reasons.

I believe the goal at this time is to provide the IS-IS WG with the
opportunity to review the changes to the IS-IS protocol which are
required by TRILL. A complete review requires a document which both
provides the normative definition of the changes and definition of the
encoding used in support of those changes. This most logically belongs
in a single document. If the consensus is that two documents be used
then so be it - but this does not mean that it makes sense to progress
the two documents independently.

In regards to the history you present below, I will only say that an
expectation was explicitly set as far back as draft-ward-l2isis-xx (back
in 2005!!) that all extensions to the IS-IS protocol in support of the
L2 technologies would be defined in an IS-IS WG draft. In more recent
times we had draft-ietf-isis-layer2-xx and now L2 technology specific
documents. My assumption has always been that once the various L2
technologies had reached a level of maturity that they would add the
normative description of changes to the IS-IS protocol processes to an
IS-IS document along with the TLV definitions as this has been the
normal practice with all other protocol extensions as noted above.

   Les

> -----Original Message-----
> From: Erik Nordmark [mailto:[e-mail]]
> Sent: Wednesday, November 10, 2010 6:59 PM
> To: Les Ginsberg (ginsberg)
> Cc: Ralph Droms; [e-mail]; isis list mailing; int-
> [e-mail]; [e-mail]
> Subject: Re: [Isis-wg] Progressing draft-ietf-isis-trill
> 
> On 11/10/10 05:38 AM, Les Ginsberg (ginsberg) wrote:
> > I am struggling to appreciate the rationale behind this two document
> > approach.
> 
> I think we are doing this in the interest of protocol document
quality,
> and bending the process rules in order to reach the desired document
> quality.
> 
> We last called draft-ietf-trill-rbridge protocol-10 first in the TRILL
> WG on Fri, 07 Nov 2008 and that last call email was forwarded to the
> ISIS WG mailing list ten days later. (The MTU PDUs where not in place
> back then, but the adjacency formation hasn't changed since then.)
That
> resulted in a few comments, but the concerns about the adjancency
state
> machine were not raised at that point in time.
> 
> We had a TRILL WG last call draft-ietf-trill-rbridge-protocol-13.txt
on
> 06/30/09. After some edits an IETF last called was done on -14 on
> 2009-12-18.
> 
> To really make sure that is IS-IS experts have reviewed the document,
a
> separate ISIS WG last call was requested by routing AD Ross Callon.
> Thus
> it was last called in IS-IS on 01/28/10. While there isn't any process
> document which describes such an extra last call, it made some sense
to
> ensure that the document was reviewed by ISIS experts that do not
> participate in the TRILL WG. That ISIS WG last call didn't result in
> any
> comments AFAIK.
> 
> Apparently that was not sufficient to get proper attention, since the
> concerns about the adjacency state machine were not raised until the
> code point document (draft-ietf-isis-trill) was last called.
> 
> So I agree things are quite unusual for this document; I think our
> desire to accept very late comments on draft-ietf-trill-rbridge-
> protocol
> is far beyond what I've seen for any other document in any other WG.
> These comments were submitted more than 6 months after the IESG had
> approved the document.
> 
> I'm quite concerned that we are setting a bad precedent by bending
over
> backwards to consider such extremely late comments, because it
> encourages suboptimal review timing. Yet I accepted the AD way forward
> to create a separate document with the adjacency state machine. I'm
> starting to think that was a mistake and I should instead have filed
an
> appeal against the ISIS WG not approving draft-ietf-isis-trill.
> 
> Sigh.
> 
>     Erik
> 
> 
> 
> > It is as if when defining the Three-Way-Handshake (RFC 5303) it was
> > decided that there would be two documents - one to define the state
> > machine and provide exposition - and a second to define the encoding
> of
> > the TLV. It could be done this way (though I am unclear on the
> > motivation) - but if so there would be interdependence between the
> two
> > documents and it would not be possible to conduct a complete review
> of
> > one without the other.
> >
> > I have no wish to extensively discuss "process" - so if there is
> > consensus that in this case we should have two documents rather than
> one
> > - I will not argue the point. But I do not see that it is logical to
> > review/progress draft-ietf-isis-trill independent of the proposed
> > draft-ietf-trill-adj (or vice versa).
> >
> >     Les
> >
> >
> >> -----Original Message-----
> >> From: [e-mail] [mailto:[e-mail]] On
> >> Behalf Of Ralph Droms
> >> Sent: Sunday, November 07, 2010 5:27 PM
> >> To: [e-mail]; isis list mailing
> >> Cc: [e-mail]; [e-mail]
> >> Subject: [Isis-wg] Progressing draft-ietf-isis-trill
> >>
> >>
> >> Ralph and Stewart (as ADs responsible for TRILL and ISIS WGs) have
> >> agreed on the following way forward with regard to
> >> draft-ietf-isis-trill.
> >>
> >> draft-ietf-isis-trill will include a normative reference to a new
> >> document to be produced by the TRILL WG. For the purposes of this
> >> email we will call it draft-ietf-trill-adj.
> >>
> >> draft-ietf-isis-trill will now proceed to the IESG for review and
> >> approval in the normal process, but will wait in the RFC Editor's
> >> queue on the normative reference (as it would wait for the
> publication
> >> of any other normative reference) to draft-ietf-trill-adj.
> >>
> >> draft-ietf-trill-adj will provide a state machine description of
the
> >> adjacency formation including:
> >>
> >> - Adjacency formation
> >> - DRB (aka DIS) election
> >> - rules for two-way and MTU matching for advertisements
> >> - creation/use of pseudo-nodes
> >>
> >> Stewart and Ralph
> >> _______________________________________________
> >> Isis-wg mailing list
> >> [e-mail]
> >> https://www.ietf.org/mailman/listinfo/isis-wg
> > _______________________________________________
> > Isis-wg mailing list
> > [e-mail]
> > https://www.ietf.org/mailman/listinfo/isis-wg


_______________________________________________
rbridge mailing list
[e-mail]
http://mailman.postel.org/mailman/listinfo/rbridge