Is anyone using XML instead of HTML for pages of content?
Yes and no. Did that about 5 years ago, still can't do it for
widespread use on a public site, owing to the limited number of user
agents that can handle it.
It may have intranet uses, but for public access I think you either
need to avoid it, do it server-side, or browser-sniff and behave
accordingly.
There's also the problem that page load is ugly on a big page. It's a
white screen until the XML document is down, the XSLT is down, and the
transform has been run. If you must do this sort of client-side XSLT,
then use the "data island" approach and client-side JavaScript / DOM
that shows a plain HTML page with a "Loading..." banner, loads the XML
& XSLT as either linked or embedded resources, then transforms one
with the other and writes it into a HTML <div>. Still browser
dependent, but it's a better implementation.
These days almost everything I write is server-side XSLT generated. If
I'm doing some complex data-navigation example (clickable table
columns to sort, that sort of thing) then I might use XSLT on the
client.
Even better is the combined approach. Write your document as _both_
XML and HTML - write it as XHTML, but in such a way that it also
contains all the data that the original XML document contained
(qualified <span>s will be helpful). Now a non-XSLT browser renders it
as the page you forst wrote, but an XSLT-capable browser has the
chance to use this as an XML source document, hide it from display,
and re-create a new display version with whatever processing you wish.
That said, there are sometimes uses for XML+XSLT directly to the
browser - when you're writing XML that isn't targetted at a browser or
human, but you need to view it for debugging. This week I'm writing
PartnerML for mobile phones. An ugly XML-based metalanguage (the phone
network stack later transcodes this into either XHTML-MP or WML, as
needed by the handset). It's a nightmare to debug, because you have to
run it through the network stack to get it into a viewable fashion.
Instead, my server does browser-sniffing and if it recognises the user
as my dev desktop running IE, it sends me an XSLT stylesheet as well
that emulates the network stack and a phone browser.
Pure XML + CSS is useless. No ability to generate elements, manipulate
atrributes, to transform ordering, or to duplicate existing elements.
Use XSLT instead.
