On Tue, Aug 13, 2002 at 12:52:25PM -0400, Greg Ames sent those random bytes:
> in the html. I am curious to hear what the W3C Validator people say.
Well, my message to W3C generated a thread of ten emails.
This is a short report of their toughts.
1 - There is no need to specify a meta charset in HTML documents if the
charset is given in the Content-Type header.
<Liam Quinn>
But there may be an additional complication: Some 404s may be in
other encodings than iso-8859-1. In that case, the header would
be wrong. As long as this is just for the built-in 'last resort'
error message that doesn't change, it's okay. But in case it's
tagged onto any arbitrary error message, it's a problem.
(So with Greg's fix Apache should be fine - Carlo)
<Martin Duerst>
BTW, a related problem is the directive 'AddDefaultCharset'.
This adds a 'charset' parameter to *every* Content-Type that
doesn't already have one. This means that if you have some
gifs, they get served as Content-Type: image/gif; charset=foo.
This is of quite useless.
<Martin Duerst>
(About the AddDefaultCharset problem noted by Duerst)
The Apache documentation implies that, but it isn't actually the case in
my testing with Apache 1.3.26. The charset parameter only seems to be
added for text/html and text/plain. It's not added for image/* or
text/vnd.wap.wml.
<Liam Quinn>
2 - About the default HTML code provided for a 404:
(Apache developers) should change <hr /> to <hr>. <hr /> is for
XHTML/XML only, but they've specified HTML 2.0.
<Liam Quinn>
3 - Some of the W3C people thinks having an option 'validate error messages' in
the validator form is a good idea, because they want to be able to validate
all html.
--
Carlo Perassi - http://www.linux.it/~carlo/
Do only what only you can do (Edsger Wybe Dijkstra: 1930-2002)
