[Wikipedia-l] Wikipedia pages DTD

Mon Dec 30 21:55:35 UTC 2002

(Richard, you may be interested in the wikitech-l development list; see
http://www.wikipedia.org/mailman/listinfo/wikitech-l )

On lun, 2002-12-30 at 11:02, Richard Grevers wrote:
> My apologies if this has been discussed before, but I just noticed that 
> Wikipedia pages carry a doctype declaration of
> <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
> 
> Since this does not include the URL to a full DTD, the curse of doctype 
> sniffing sees most browsers render the page in "Quirks" or "Bugwards 
> compatible" mode rather than "strict" mode.

Yes, this is intentional.

> In the interests of making 
> wikipedia accessible, I think it would be better to force strict mode - 
> then editors would be more likely to notice markup errors which affect the 
> rendering on browsers which do not correct for bad markup.

We don't use the strict DTD (or, better yet, XHTML) because our
hacked-together wikicode->HTML parser currently can't guarantee that it
will generate well-formed output (particularly if there is raw HTML in
the page, which is munged a bit but not always correctly). If
well-behaved web browsers reject the page or massively break page
rendering due to a minor error, it's going to be mighty difficult for
editors using them to click 'edit' and try to work around the problem!

(On a similar note, nested HTML tables without closing </td> and </tr>
tags, while acceptable under HTML 4 standards, break Netscape 4.x, even
to the point of *crashing* it outright. This cropped up a while ago on
some of the Canadian provinces articles; the person who noticed the
problem couldn't edit the pages to look for the problem, since the
browser crashed before an edit link was made available. Really strict
DTD parsing isn't quite as bad -- it should at least tell you the
problem! -- but makes it hard to continue browsing.)

Volunteers for a new XHTML-safe parser are welcome...

-- brion vibber (brion @ pobox.com)