Re: initial issues

[Thomas Narten <narten@raleigh.ibm.com>]

Hi Ben.

> As an aside, I personally think a GSE-like approach is more realistic
> given the views on multihoming I've seen expressed on various lists and
> at the multi6 BoF.  Why all forms of GSE were rejected in favor of the
> obviously untenable, unsolved current state is simply beyond me.

Let me jump in here since this is another one of those potential hot
buttons. Warning: I'm certainly one of those viewed as rejecting the
approach and being an "obstructionist" with regards to continuing work
on the topic (or so I am told). So my view should be considered
biased.

The folks that looked at GSE back in 1997, concluded that there was
too much risk in adopting the full proposal at that time. This was a
unanimous view of the workshop participants (see
http://playground.sun.com/pub/ipng/html/minutes/ipng-minutes-feb97.txt
for instance).  To simplify, is that it was felt that there were some
open issues that were not fully understood and more work was needed to
be done. But the participants didn't want to take on that work
themselves either. 

A number of folks that I talk to that participated in that original
discussion believe that GSE is not a panacea that solves all
problems. Rather, it moves some deep and complex problems to other
places in the overall architecture, where they may or may not be any
easier to address. (Note the emphasis on may or may not be -- the
point is that it's not entirely clear or immediately obvious, at least
to some.).

Unfortunately, what has happened since then is that the two sides of
the issue have apparently become VERY polarized and seemingly have
moved VERY far apart from each other with a huge gulf between them and
very little real technical discussion. I.e., to oversimply, some folks
seem say to say something like "GSE trivially solves problem X, why
don't we just do it" (where X can be multihoming, renumbering or
something else), while anyone else who seems to question such an
assertion risks being painted as an obstructionist that just wants to
kill a Good Idea.

Someone else has already pointed out the gse-analysis paper. Well, its
viewed by some as seriously flawed. For better or worse, the authors
have pretty much given up trying to work on it. It's been a long time
since the proposal was first discussed, and its hard to get excited
about and find time to work on a document that one only gets flamed
for.

So, personally, I do think there needs to be a venue for having a
*serious* *technical* discussion about GSE, because I think some
unrealistic expectations have been placed on it. But until there is a
real technical discussion, all we'll likely do is start a flame war. I
don't believe this list is the appropriate venue for such a discusion,
at least not at this time.

So for now, I think it would be good to stick with your first note,
i.e., defining just what multihoming is and what its requirements are.

Thomas