RE: peace and quiet

Note that Literal-and-discourage was the conclusion that the Plenary
Group's straw poll reached, for more or less the reasons we've seen in the
previous notes under this subject. This solution is something that both
those who think relative syntax is evil, and those who understand the
objections but are already using it, can live with.


If it helps, think of Literal-and-discourage as really being Forbid with a
migration path. We're recommending that folks use other mechanisms (which
we've discussed) to bind the namespace name to a relative URI reference if
that's what's desired. But we are allowing the relative syntax via a
least-effort solution in order to give folks a reasonable amount of time to
update their documents and tools.

(The actual proposal for "discourage" was something closer to "permanantly
deprecate" -- meaning it really isn't supported, but we may never get
around to checking for it, so folks who completely forget/refuse to fix
their systems won't have them fail any time soon.)


Despite TimBL's expressed doubts, I believe that this conclusion -- which
followed a discussion just about as intense and deep as this one -- really
did reflect a deep understanding of all the conflicting goals and an active
attempt to find the best possible balance between them. It supports all the
cited use-cases (though it requires indirection via explicitl bindings for
some of the ones which weren't within the scope of the original Namespace
spec). And it gives us a direction for the future without breaking existing
practice, at the slight cost that those who really want to absolutize the
name have to accept responsibility for doing so explicitly.

Would we have chosen differently if we were designing Namespaces from
scratch? Probably; I suspect that in that case the name would have been a
URI and relative syntax would never have been permitted. But then, the DOM
might also be significantly different if we we were designing it today
without concern for ease of migration from "DOM Level 0" (pre-DOM browser
scripting). Once something has been in use for a while, it's sorta impolite
to the users to change it more than absolutely necessary.

If we really think that the Literal solution is a long-term problem,
consider writing and releasing a Namespaces 2.0 spec which closes that gap;
it could make something like xmlns-binding: official at the same time.
(Probably in a more sophisticated form than my basic proof-of-concept
sketch.) But Namespaces 1.0 is what it is, and Literal/discourage lets
everyone live with it until someone is willing to try to tackle 2.0.

______________________________________
Joe Kesselman  / IBM Research

Received on Wednesday, 24 May 2000 10:13:02 UTC