>Your point is well taken, however, in my view, the most authoritative
>source for information about a document is the document itself, hence
>the order of perferences that I list.
Define document for me, and I may agree. I think that a "document
object" is comprised of "the informational unit and associated meta
data". I do not see that the meta data should, or even can reside
inside the informational unit.
A good example of the distinction I'm trying to draw is between
attributed, and non-attributed file systems. Attributed file systems
allow meta-data to become part of the object *without* being *in* the
object itself. In this repect, I think most people will agree that the
type system of the Mac filesyetem is far better than the DOS
>I have a hard time seeing WebMasters configuring their site to send
>the proper charset information. The servers aren't there, tools
>aren't there, and WebMasters' time is a real premium.
Well, Apache gives a method (the *.asis file) for doing this, and I
proposed something like it (the *.mim) type as a simple way of doing
Still, this is a good point. The current tools are broken, and need to
be fixed. I have ranted on about this for quite some time now, and get
a feeling that most people are apathetic at best.
>I say: let the document authors worry about supplying the proper
>charset information. The way to do this is by using the <META> tag.
>Hack or not, it is a good idea.
Anything that is unreliable and hampers adoption of more robust
solutions is not a good idea.
>In any case, this argument is somewhat moot, since I've never
>experienced getting charset information in an HTTP header. In our
>implementation we can go either way.
You will. I think that the first round of HTTP 1.1 servers will be
much better at respecting the standard.