Re: Version identifier in URL
Christopher Seiwald (seiwald@p3.com)
Wed, 29 May 1996 21:32:18 -0700
Dan Connolly said:
| It's OK for individual systems to have conventions about
| the syntax of HTTP URLs, but it's an abstraction violation
| to standardize such conventions across all HTTP servers.
Then I said:
| I'm not sure I understand Dan Connolly's objection to standardizing
| a version extension to URLs, other than that he is objecting to it.
Then Dan, wheeling about, retorted:
| URLs are supposed to be opaque strings that identify resources.
|
| Information providers may encode semantic information in them, but
| clients must treat them as opaque, except by private agreement.
|
| [ Later, guns ablazing ]
|
| I assure you that Tim Berners-Lee and myself will do everything we
| can to prevent you from winning.
Now that I understand this position, I agree. Since the major point of
version-decorated URLs is to avoid requiring client awareness, leaving
URLs opaque is logical.
I believe what we are discussing, then, is a convention for attaching
version information to URLs, rather than a standard. There are many
such conventions -- hosts named "www", for example -- that make URLs
readable to humans, even if they do nothing for servers and clients.
It seems to me that if we do promote version-decorated URLs, we should
propose a convention. This was the thrust of the ,;/ discussion.
The whole topic of version-decorated URLs is, however, somewhat at odds
with Dan's talk of versioning with links. Version-decorated URLs
presuppose that the server has external version control over the web
content. The use of links, it appears, is an attempt to implement a
versioning system _in HTML_. Before I slight this approach, maybe Dan
could explain how those links get there: humans type them in? created
by web authoring tools? automatically inserted by an external versioning
system on document retrieval?
Christopher
----
Christopher Seiwald P3 Software http://www.p3.com
seiwald@p3.com f-f-f-fast SCM 1-510-865-8720