Re: draft-ietf-conneg-media-features-02.txt; LAST CALL

[Al Gilman <asgilman@xxxxxxxxxxxxxxxx>]

to follow up on what Graham Klyne said:

> >One could argue that having three tags which are non-orthogonal would
> >be more difficult on negotiation protocols.  However, if you are going
> >to make an extensible negotiation protocol coupled to an essentially
> >open registry like the feature tag registry, you will have to deal
> >with non-orthogonality anyway, because sometime, probably very soon,
> >someone is going to register a tag that is non-orthogonal to other
> >existing tags.
> 
> In the current version, there is no way to deal with non-orthogonality
> _within_ the capability description framework.  Such mechanisms were
> deliberately excluded in the interests of simplicity, and ensuring a
> tractable feature matching algorithm.  It may be that future developments
> will address this issue, but I'm not assuming that will be so.
> 
> External ad-hoc measures can be employed, and I expect that
> application-specific schema[ta] will play a part here.

This is what I have been trying to warn about.  The registration
facility lacks the tools to enforce orthogonality, and even a
plan for how to bind the wounds.  There needs to be sufficient
means for a third registration to delare relations (assertions,
cross-propositions) which bind tokens introduced in first- and
second- un-coordinated registrations.

The present drafts, as I vaguely understand them, lay no
foundations for this but allow annexes in registrations so that
there can be an after-the-fact rationalization across diverse
registrations.

In other words, if I submit a registration with some new terms
that includes a schema documenting the relationship of these
terms to earlier-registered terms, the registry will house but
not comprehend the relationships asserted in the schema.

Al