Keith's Weblog: Comments on "Understanding charsets on the web"

by Sam Ruby

Fri, 11 Feb 2005 18:23:28 +0000

I wish things were so clear. In fact, the W3C recommends otherwise for XML.

And when you factor in web browsers that are broken, you end up doing like http://www.google.com/ does, which is to omit the charset for HTML and put it in the meta tag.

by Keith

Fri, 11 Feb 2005 19:19:05 +0000

Well, that makes sense too. HTTP has primacy but if you don't specify the charset at the protocol level it defers to the content. Though, I didn't know that they recommend not using text/xml for XML... so, it seems we should all be using application/xml by default .

As for how Google does it -- that doesn't bother me too much since we're not dealing with logical inconsistencies, merely implementation bugs. But the W3C's recommendations for XML presumably apply to XHTML as well, so I suppose that implies we should all be specifying the charset only in the (X)HTML anyway?

by Anne

Fri, 11 Feb 2005 20:31:08 +0000

What is wrong with using application/xml by default? text/xml has issues and will be deprecated eventually.

Those W3C recommendations seem a bit weird. Not sure why they do that. I'm also not sure since when the W3C is the authority to advice people on what to do on the HTTP level.

by Keith

Sat, 12 Feb 2005 00:12:13 +0000

Nothing's wrong with application/xml by default. It's just that it's like, "Now you're telling us?". If text/xml is no good I'd have preferred they never created it. Now I have to change my code, and what's worse is that I don't know if there are any compatibility issues between text/xml vs. application/xml. For example, will anyone have a problem if I start serving my RSS feed as application/xml instead of text/xml?^[1] So, because I don't know, I'll have to check accept headers for application/xml and serve text/xml as a fallback.

Plus, if text/xml has problems and will be deprecated eventually, why isn't it deprecated now? That's what deprecated means. It doesn't make sense for anything to be in the position of "going to be deprecated".

Footnotes:
[1]: And XML-RPC? Though, XML-RPC is always POST, so maybe the issues with text/xml don't matter?

by Anne

Sat, 12 Feb 2005 10:48:06 +0000

There are no problems. application/xml is accepted as well. It just is that text/xml has a default charset of US-ASCII and application/xml does not. (Default is UTF-8.)