Re: Proposed announcement to rest of IETF

[Vern Paxson <vern@ee.lbl.gov>]

> > Costa, without requirements analysis you will not get anywhere in the IETF.
> 
> Can you point me to some documents that indicate what goes into
> a requirements analysis? I'm fuzzy on the concept.

I don't think we have such a document per se, but what goes into a
requirements analysis is something like:

	1.  Here's the problem [domain] we're trying to solve
	2.  Solutions for this problem must do X because of ABC.
	3.  Solutions for this problem must do Y because of DEF.
	4.  Note that doing both X and Y means the solution must
	    also do Z.
	5.  It's out of scope for solutions to do Q.
	6.  It's optional for solutions to do W.
	7.  Note that one possible approach is Foo, which has the
	    following additional consideration: ...
	8.  There's also the Bar technology, but it's geared for a
	    somewhat different problem domain, because of blah-blah-blah.

These last two aren't strictly appropriate for a requirements document,
but if done in objective terms can be natural discussions to include.

> The goal is not to replace NFS. In fact, some of the proposals out of
> this proposed working group may actually help NFS. Again, though, the
> principle goal is to make SCSI work over IP networks.

I think Brian's point is that until the requirements are nailed down, there
will be confusion over exactly what problem is it we're trying to solve.
In particular, I believe the problem is not "run SCSI over IP" but something
a bit more abstract than that, for which one of the possible solutions will
be "run SCSI over TCP over IP, with the following tweaks to the SCSI spec,
and the following TCP options, and ...".  And we want to understand how
that abstract goal does indeed differ from the abstract goal that NFS attempts
to address.

		Vern