First of all I want to state my position. I don't like extensions. I
want URLs to identify resources and not a specific file format. I guess
according to Ruby I'm a purist. :)

David M Johnson wrote:
> 
> On May 3, 2006, at 4:49 PM, Allen Gilliland wrote:
>> I know what you are saying and part of my mind wants to agree, but
>> ultimately I still have to say "yes" .html is an implementation
>> detail.  A url is meant to point to a resource and that's it, how that
>> resource is defined is an implementation detail.
>>
>> Also, there are plenty of reasons why a single resource (URL) should
>> be available in multiple content types.  That is seldom used these
>> days, but it makes a lot of sense.  Along the same lines there is the
>> somewhat crazy example of, "what if XX years from now you want that
>> same url to return something other than HTML?"
> 
> Right. The client sends an HTTP accepts header and the sever uses that
> to determine what to send back. Content negotiation.
> 
> I just exchanged a couple of emails with Sam Ruby. He likes the idea of
> extensions and uses them in his blogging software, but only to indicate
> content type. He says that web purists argue that extensions are not
> necessary, but that content negotiation doesn't work in practice.
> 
> Perhaps in XX years, when we're doing what you suggest, content
> negotiation will have been fixed.

I don't think we need to wait XX years. Which URLs patterns would be
broken today if we don't use an extension? I think the common cases are
well handled on the web (text/html, text/xml, etc), if not Content
negotiation would be dead already. Of course, RSS will always be a
problem, but I think it's an issue, the readers just dealt with it. Ian
[1] says it's dead, but in fact he's only making a suggestion to
retiring it.

[1] http://ln.hixie.ch/?start=1144794177&count=1

> 
> And BTW, I'm not arguing for extensions here, just adding a data point
> -- I really don't know what the right decision is.
> 
> - Dave
> 
> 

Reply via email to