Being a tinkerer and technophile, or more appropriately, a "simplicity/cleanliness in implementation/interface"-ophile, I've been trying to get a trivial portion of my site to serve as
application/xhtml+xml to the proper browsers, while allowing me to use
URLs without file-type extensions (which I think are a little
ugly anyways). I ran into plenty of trouble finding a solution which IE agrees with, so I'm happy to share it to those who are interested.
There are loads of
purported solutions on-line, such as writing the
HTTP Content-type header with php or python, or using Apache's mod-rewrite to redirect to the proper page based on the browser's
Accept header. For me, the scripting solutions are out, as I'm interested in doing a bit more than just serving the same XHTML 1.0 in different ways. For instance, I'm using XHTML 1.1 now, and plan to use XHTML 2.0 in the future (when IE probably still won't support
application/xhtml+xml). I don't want to
just have the page interpreted differently, I want to use the new language to make my markup more simple and semantic.
This leaves the option of maintaining two copies of the files in question and having the server determine which to serve when a request for an extension-less file such "index" is sent. There are two options here. The first is
mod-rewrite, which unfortunately I couldn't get to work, and without access to my friend's "apache.conf", I can't debug. The other way is called "Multiviews", it's used primarily to serve different translations of a given page based on the browser or user-agent's language preference. This works unambiguously because basically every browser gets the
Accept-Language header right. Unfortunately the
Accept header is more problematic. The problem here, as usual lies in the most recent version of IE, which doesn't list much of anything in it's
Accept header. For example, rather than list
text/html as the preferred document type along with others IE accepts, it includes the blanket
*/*. This will of course match with any and all MIME types and therefore give no information about what IE "likes". So the trivial solution, which involves giving the browser what it asks for, is out. Time for a
Hack.
If one just does as he expects he needs to do in this situation, and puts the two files up as normal, turns on
multiviews and trys to access the page in Firefox and then in IE, they'll find that the
application/xhtml+xml is preferred in both. Unfortunately since IE won't interpret the page, users of IE (who really aren't bad people) get a download dialog box. This won't do, so I did some tinkering and I finally found a solution which actually works.
I present to you my awesomely bad hack:
Append white-space to the end of the smaller file, such that they are equal in size.
The same XHTML is about 5-10% smaller, keeping in mind the need for IE-specific stylesheets and alerts. Lacking other reasons, Apache is providing the XHTML to (the seemingly indifferent)
IE due to the smaller file size. The
Multiviews algorithm spells this out. Seemingly there should be work-arounds, using .var and quality-of-source values, but I was unable to get this to work. The white-space has been the only workable solution I've come up with so far is . This results in the following behavior:
application/xhtml+xml
text/html
Once I script the white-space addition, serving both will be a simple process. If you have a better solution, or problems with the one above, feel free to comment or contact me.
Edit: Level to the Rescue
So here we have open-source web-serving. Someone has provided me a much better solution in the comments below, which I'll discuss now, lest we all do things without understanding them.
The approach is to use the Apache
level designator to force
IE to prefer HTML over XHTML. One can see how this works in the
Apache documentation for the Content Negotiation algoritm.
I'd done some experimenting with the
qs factor, but it had affected multiple browsers, rather than just one. Referring to
the algorithm documentation, all quality-of-source values being equal, any browser which advertises that it prefers
application/xhtml+xml over
text/html in its
HTTP accept-header will have the algorithm settled in step 1 of the algoritm. This covers Firefox and Opera. The others,
IE & Safari, give both
text/html and
* a quality factor of 1, and so continue on down the steps. They had been finishing at step 8, which decides based on content-length. The much nicer solution is to stop them at step 4 in the algorithm, this time on the level parameter.
A big thanks to Chris for the tip.
Edit 2: Not so fast
Alright so level isn't working for me...
IE still tries to load the
application/xhtml+xml files if they're of smaller file size. Tried level in both the .htaccess and through type-maps to no avail. If anyone has this working or another alternative, I'd be happy to hear it...