Re: FW: fragment prose proposal

> You guys are gonna have to educate me. If (as you seem to be 
> asserting) one can plug in any old retrieval mechanism for a 
> self-declared non-dereferenceable scheme (such as INFO), then 
> presumably one could do the same wrt any other scheme - e.g. construct 
> an FTP retrieval mechanism against an HTTP URI, or a Handle retrieval 
> mechanism, or whatever.

Yes, that has always been true.  It was the original purpose of HTTP
proxies, in fact, and the way that most www libraries work.

> Am I missing something here or don't things begin to fall apart if 
> just any ad hoc retrieval mechanisms are recognized client side? What 
> is the point then of a URI scheme if clients can override any given 
> protocol and do as they want? Isn't that just a smidgen close to 
> anarchy?

The URI scheme does not denote a protocol.  It never has.  It defines
a mechanism of name assignment.  One of the easiest and most useful
ways to assign names is via information systems that use network
protocols.

> BTW - Isn't it marvelous that absolutlely /no/ browsers can recognize 
> a generic URI string, even now some ten years after? :~) All of 'em 
> just manage to scrape through with a few hard-wired schemes (the 
> 'GETables') which they deign to recognize. Quite surprised that even 
> now the standards-compliant-wannabe clients simply cannot recognize a 
> URI string in context. What does that say about the state of the Web?

It says that browser developers don't read design specs, or simply
don't care enough about URI extensibility as opposed to bugwards
compatibility.  The open source applications don't seem to have any
of the problems you describe.

....Roy

Received on Saturday, 28 February 2004 14:44:19 UTC