On Mon, 16 Apr 2001, Murray Altheim wrote:

> "Charles F. Munat" wrote:
> >
> > Sean B. Palmer wrote:
> > "Thus, I guess I conclude that embedding information into XHTML is
> > quite a good idea, when you weigh up all sides of the debate."
> >
> > I disagree.
> >
> > Yes, it would be good to have a way to embed metadata right in the XHTML,
> > but for the near future it would be better to link to a separate metadata
> > document. Here's why I think this:
> >
> > The vast majority of page loads for at least the foreseeable future will be
> > by legacy browsers unable to do anything with the metadata. If the metadata
> > is embedded in the XHTML, all of those page loads will have to bring down
> > the (useless) metadata. What a tremendous waste of bandwidth.
> >
> > If we can link to the metadata via a link element, then only those user
> > agents needing the metadata will access it. It will still be accessible to
> > anyone using older browsers (though they might have to look at the source to
> > find the link), but we'll save a lot of bandwidth.
> >
> > I would think that this would be of concern to every site developer. Making
> > it possible to link to metadata might encourage its early adoption.
> >
> > So first, let's find a mechanism to link to RDF from XHTML. Then we can find
> > a way to embed it at our leisure.
> The difficulty with external links is first of all management. If you have
> 1000 documents you'd have 1000 metadata files. It's a lot simpler to add
> the markup into the document, a lot easier to be sure you've got the right
> metadata, and spiders can obtain the data directly. I also think it'll be
> a lot easier for authors and braindead tools. Documents will also remain
> rather portable.

It doesn't follow from the use of LINK REL that you have to manage
metadata on a file-per-document basis. You might point into a CGI or
servlet (hopefully disguising this to the outside world), or autogenerate
those files from a database managed elsewhere. I do share your bias
towards (at least some) embedded metadata though.

> The argument on bandwidth seems a bit unwarranted. The amount of metadata
> is usually pretty small, much smaller than the smallest GIF image on a
> web page. For example, your message in my mailbox took up 1221 characters,
> which would probably be three or four times (at least) the size of the
> metadata in a typical XHTML document (a Dublin Core record is not very
> big). Most GIFs are at least ten times that size.
> This is not to say that somebody shouldn't harass the HTML WG about adding
> in a standard feature in XHTML 2.0 to link to DC metadata (or a general
> metadata link with attribute stating a notation type of "DC").

Yes, pointing to metadata generically isn't so informative. It would be
nice to have some mechanism for hinting at the sort of thing found at the
other end... (hmm... doesn't XLink do this for us, allegedly.)


Received on Monday, 16 April 2001 19:28:16 UTC