Musings

Did you spike the “survey” link with that j before the protocol for effect?

I wonder how long it will be before those of us with comments enabled think twice about linking to a weblog post without comments, knowing that we will become the comment host for that entry, for our readers.

You are certainly the only person I know of with a evident need for XHTML. However, there are also uses for XHTML that aren’t apparent at first glance. I switched over because I neeeded to do some self-scraping, and doing it with an XML parser seemed saner than using a huge chain of regexes or writing my own SGML parser.

I’d also quibble with his third test: the spec doesn’t say MUST use application/xhtml+xml, it says SHOULD. The word the isn’t normative, SHOULD is. MUST means must, SHOULD means should know why not and what it will break. I don’t use application/xhtml+xml because the gains are minimal, and since I’m not using a real XHTML tool to produce my pages, there’s the risk of delivering Mozilla’s XML error page between the time that I save a screwed up entry, or someone leaves a comment that gets past Sanitize, and the time I see it and fix it. By delivering XHTML as text/html, I get the safety net of browser error handling while still getting the benefits of XML for myself. Hixie’s understandably focused on browsers and nothing but, but that’s not why I need or want XHTML.

On “spiking the survey”: No, Jacques’s site was not included. :)

As for the “SHOULD” versus “MUST” – yup, I recognize that. And I understand the need to serve up pages as text/html to XHTML-unaware browsers. That’s why Test #3 simply checks whether you’re serving up the right MIME-type to a XHTML-aware agent, namely Mozilla. My feeling is that if you’re advanced enough to use XHTML, you’re advanced enough to add a couple of lines to your .htaccess file.

But okay, yes – I suppose one can quibble with test #3. There is a lot of confusion out there on this issue, and the unhappy state of browser compliance is not helping matters. Let’s say for the sake of argument that I completely drop test #3. In that case, what Jacques says still stands: A) even on pure validation, compliance is pretty miserable (about 10%, among the Alpha Geeks) and B) the point of my exercise is really to look ahead to XHTML2, where the “SHOULD” becomes a “MUST”. At least badly-formed, improperly MIME-typed XHTML1 can fall back into tag-soup HTML. XHTML2 is when it starts getting scary.