Clarified and expanded on the fact that the XML Declaration and XML-stylesheet reference (along with the DOCTYPE) are part of the XML Prolog - and how to write each part.

Clarified that ' is unsupported in HTML 4.

Reworked HTML comments in scripts.

Added Point 11 in "Doing the Right Thing" section.

My thanks to Lachlan Hunt, Jukka K. Korpela, Ben Millard and Torsten (above and forum member at Accessify.com)

As I say, I'm not an expert, but I am rather surprised that it works.

The accept header for Firefox 1.07 under GNU/Linux is as follows -

text/xml,application/xml,application/xhtml+xml,text/html;q=0.9,text/plain;q=0.8,image/png,*/*;q=0.5

...your regular expression doesn't match anything in that string. By my understanding your script would therefore serve text/html to Firefox when, according to its accept header anyway, it prefers application/xhtml+xml.

I just had a look at Simon's article and his reg exp seems much better. It is in fact very similar to the one I use myself:

$htmlmimetype = '/text\/html(?:;q=([01](?:\.\d{1,3})?))?/i';
$xhtmlmimetype = '/application\/xhtml\+xml(?:;q=([01](?:\.\d{1,3})?))?/i';

...so I would say that wouldn't I ;o)

Torsten: I can't recall my exact thinking behind using that regex in that way, but what I can say is that I've been using that script (or versions of it) for about two years now, and I haven't had it blow up in my face yet.

I've just had a quick play with Firefox, changing "network.http.accept.default" (vie about:config) to contain various different q-values and I couldn't get the script to do anything unexpected. However I am not convinced that you don't (whoa, too many negatives there) have a point, so I'm going to go back and have a proper look at the script later in the week.