Hi Martin and all,
see below...
Martin Stiemerling wrote:
However, in the full spirit of progressing NSIS to RFC status and not
to abolish all the excellent work of the past, the decision to go for
experimental looks natural in the light of the dead lock.
Remember, that GIST is only the first document out of the whole
series that is still to come (e.g., QoS NSLP, NATFW NSLP, QSPEC,
etc).
Please note, that downgrading to experimental, also likely applies to
the NSLPs.
Please let the WG know your opinion about that!
I'm really disappointed. IMHO the WG has done all necessary
steps to make it a really good protocol. If we didn't care about
protocol quality, we could have finished GIST long time ago.
It's not the WGs fault that using Router-Alert turned into a
potential/real show stopper later on, though it is still used for RSVP.
We asked all the experts at the time, and, it was the recommended
practice to efficiently intercept signaling packets.
Therefore, it really is totally weird if one reads Adrian Farrel's
DISCUSS POINT 3 on whether the currently RAO-less option is
efficient and fast enough (esp. if you recall that he actually
already reviewed the document some years ago).
The IESG seems now to pull into the RAO direction again(?), if
reading Cullen Jenning's comment.
I also cannot understand Ross Callon's comment why we need
GIST at all in addition to RSVP and RSVP-TE, though our WG
published various RFCs on this topic already at the very beginning.
Ron Bonica's comments are also depressing. I don't agree with
his comment on complexity at all, because we have several
interoperable implementations, which were not too hard to
code. Furthermore, if you take page count as a measure for
complexity, OSPFv3 would never have been standardized...
As an L3 control protocol for IP resources, it naturally has
to deal with a lot of things, so I don't think that there
are easier solutions for the same feature set.
However, if going for experimental gets us out of this
whole mess quickly, let's do it. Please no more delays.
It would be good having an official RFC number anyway for vendors
in order to implement it. Probably, then they see the benefits
more clearly...
Regards,
Roland