Re: Proposed Design Principles updated

On Thu, 5 Apr 2007, olivier Thereaux wrote:
> 
> Isn't the case of CSS stylesheets served as text/plain a counter-example 
> of this?
> 
> Some browsers started ignoring ill-served CSS, sites "broke"... people 
> eventually fixed their sites' config. The browsers' choice to be strict, 
> there, did not "break the web", arguably it helped fix the web.

Actually no. One browser started ignoring CSS files in one very specific 
case, namely only if the document had a DOCTYPE that, at the time, was 
basically unused. (I was one of the two people pushing for this.) It broke 
some sites, but comparatively few in the grand scheme of things. (Though 
we still get bug reports about this today, so maybe it wasn't that good an 
idea after all.)

It's not as difficult to do things like that when you don't have any 
legacy content. For example, theoretically all kinds of things can be done 
to XHTML, since you don't have any legacy content to worry about. The 
problem is in fixing things that would break existing content. For 
example, while treating text/plain files as text/plain instead of HTML 
might work ok (some browsers do it already), there's simply no way you can 
treat Video files sent with the Apache default text/plain type as 
text/plain, or treat PNGs sent as image/gif as GIF, or treat Atom files 
sent as text/html as HTML, or even look at the MIME type of files pointed 
to by <script> elements. The HTML5 spec has a list of what sniffing is 
basically required, put out in very specific terms. I think it's 
imperative for security reasons that browsers all follow the same rules, 
hence why we should have a spec that defines this.

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'

Received on Thursday, 5 April 2007 05:51:42 UTC