Re: iSCSI: Some proposed vendor-specific (X-) keys

[Luben Tuikov <luben@splentec.com>]

Paul Koning wrote:
> 
> Excerpt of message (sent 7 June 2002) by Luben Tuikov:
> > I haven't checked for those lately, (especially in the login procedure),
> > but any time you see ``MAY'' or ``may'' in the draft and a target
> > and initiator arrive at different outcomes _just_ by taking one
> > or the other route, you have ``compliant-non-interoperability''
> > (as you coined the term).
> 
> That is not true at all.  "MAY" is fine if either choice results in
> behavior that is acceptable to the other side.  In fact, that's the
> only place where a standard is allowed to use it.  Sometimes,
> achieving that requires that side A communicates its choice to side B;
> in other cases it doesn't.

That is not true at all.

Paul, you are NOT reading the whole sentence. Here it is again:

Any time you have ``MAY'' or ``may'' in the draft _AND_
a target and initiator arrive at different outcomes
just by taking one or the other route, there will
be ``compliant'' interoperability problems.

Those two messages show exactly what I have in mind:

http://www.pdl.cmu.edu/mailinglists/ips/mail/msg10158.html
http://www.pdl.cmu.edu/mailinglists/ips/mail/msg09751.html

This has since been fixed in the draft.

> A very simple example is the use of MAY in the rules for responding to
> protocol violations.  Since those cases don't occur in the first place
> in conforming implementations, neither choice can possible result in
> compliant non-interoperability.

Using this kind of ``proof'' you can easily prove anything...
 
> If the spec allows a choice -- either with MAY or with MUST -- but the
> conforming other end will for one of those two choices -- then the
> spec is broken, pure and simple.

Paul, only ``MAY'' allows a choice. ``MUST'' doesn't.
I'd rather ONLY see ``MUST'' and ``SHOULD'' in a spec.

And no, it is not so simple. Finding problems like those
involves drawing a decision graph and enumerating the final
nodes/states and then comparing if you get the same ``number''
for a target and initiator taking different route on MAY.
This is described in the literature.

-- 
Luben