Re: comments on draft-hoffman-char-lang-media

[email protected] Wed, 26 Apr 2000 10:45:46 -0800 (PST)
Newsgroups gmane.ietf.medfree
Message-ID <[email protected]>
> In a private conversation about media features and
> draft-hoffman-char-lang-media,
> I thought that it became clear that we should suggest two clarifications on
> the role of a 'language' and 'charset' tag registered in
> draft-hoffman-char-lang-media; in particular, a 'language' feature describes
> a preference (a client that can accept language=fr might also accept an
> image of a sunset that has no language, or a multilingual document, etc.)
> but that the 'charset' feature describes not only the ability to process
> the particular charset but a stronger requirement, that the data be
> delivered
> as *labelled* with the named charset. So if you want charset=us-ascii, you
> don't want charset=ibm367 (even though they're the same).

I was meaning to bring this to Paul's attention -- sorry.

I believe it would be nice (and relatively easy) to address the language
tag issue in the draft. I don't insist on it, however, if people feel
this should be dealt with in the base specification.

However, IMO the latter issue
with charset aliases has to be addressed in the draft. (This is speaking
as an AD.)

But my take on the issue is a bit different from Larry's. (No longer speaking
as an AD.) I think that the document should say that aliases SHOULD NOT be used
in media feature expressions and that feature expression manipulation tools MAY
convert aliases to the the principal name for the charset.

I guess I don't see the point of insisting on pure label match semantics
here, as it doesn't fix the potential interoperability problem. I'd rather
simply say "don't use aliases" and allow correction of that error if
it comes up.

> I think that we might want to consider revising the feature registration
> BCP to encourage registrations to be clearer about whether a feature
> represents a preference, capability, characteristic, and whether the
> feature requires explicit labelling as such.

I'm all for it if it's feasible, but my guess is that there are lots of
ratholes in the vicinity.

				Ned