xml-dev - Re: Real XML Site

[Paul Tchistopolskii <paul@qub.com>]

----- Original Message ----- 
From: Eric van der Vlist <vdv@dyomedea.com>

I realized that the real question is :
"is .xml closer to .txt or is it closer to .cgi?" 

> My purpose wasn't to say if the practice is good or bad, but to
> acknowledge it's common ;=) ...

I'm also not sure is it good or bad. 

However  ... when I'm asking GET /some.xml 
and I'm receiving *not* xml in return - this looks really 
strange... ( now, after Tim said that ;-).  

When I'm asking GET some.txt or GET some.zip or GET some.gif  - 
I'm receiving the 'corresponding' mimetype ... 

But when I'm asking for .asp or for .cgi it is actually OK to return
unpredictable mime-type ... Hmm ...

.asp or .cgi are pointing to mime-type 'unpredictable' ;-)

> > I think this practice exists only because there is no simple and clear
> > URL-rewriting layers in most of the HTTP servers I've seen.
> 
> I am not so sure. I rather think it depends on the semantic you are
> attaching to a HTTP URL.

Yes. And also it depends is it URL or URI we are talking about.
Actually, I realized that what I wanted from XT was URI support 
in standard libraries, not only URL support. This was side-effect - 
not realted to .xml actually.
 
> If you see it as a kind of FTP request and attach a meaning like "get me
> that file" then, yes, you'd expect that the extension will be coherent
> with what you get back.

URL ( URI ) identifies 'something'. Asking for HTML 
( but requesting some.xml ) looks like weird indetification of 
'something'. Or maybe not.
 
> If you see HTTP as a service request (ala CGI) then there are other and
> better means to specify the mime type(s) you are expecting back than
> relying on an extension.

.... well ... there is already some big company which relys
on extension in the situation when mime-type is suspicious ;-) I'm talking 
about MS IE. ( Not only it looks at file 'extension', but it sometimes looks 
inside the file for <?xml header ;-)

> If you see a HTTP request as "execute this service and get me some
> results back", 

Aha! I think that this is a 'special' extension/mime-type ;-) I mean that 
we can assume that "when it ends with .asp or .cgi" - we are talking to 
the service. Service is 'special mime-type'. Void * ;-)

This means that returning HTML for some.xml is still 'inconsistent'. 
It should be not some.xml but some.cgi or some.?

I think in Cocoon presentation (?) they were talking about the 
usecase :

"My friend is browsing web with lynx. I'm browsing web with MS IE. 
I'm forwarding http://some.xml to him, and he invokes the URL. Cocoon 
should invoke appropriate stylesheet, not serving  too much images 
to lynx".

Replace 'lynx' with 'some gadget that wants not HTML, but some other 
format' - and this means that this URL is actually pointing  to the service. 
Void *.

For the sake of consistency it should be http://some.cgi
( or just http://some ) Void * ;-)

> > I think situation is somehow similiar to HTML. What you ( and many
> > others, including myself ) are saying : "URI is confuizing? Big deal -
> > it works" , is in fact close to saying : "HTML is not well-formed?
> > Big deal - it works".
> 
> They are confusing as long you overload them with extra semantics ;) !

Maybe. However, I think life could be a bit easier in the world when .txt *usually* 
points to the text file. 

The question is : "is .xml closer to .txt or is it closer to  .cgi?" 

I think .xml is closer to .txt than to .cgi but I'm not sure. ;-)
 
Rgds.Paul.