Re: http URL defn in HTTP/1.0

>>I hope not, because I'm about to propose a "version" parameter to
>>allow access to back revisions of a page stored in a version
>>control system like RCS or SCCS.
>I hope nobody intends to make such a proposal until after the 1.1 
>draft is submitted...

Why?  The syntax for HTTP URLs hasn't changed between 1.0 and 1.1,
so there's no dependence on the version of HTTP.

>... in any case I think there might be somewhat more of a problem 
>with versions than one might think at first glance. Having built a 
>server on top of CMS a couple of years ago I don't think that there
>is a comprehensive version naming strategy avaliable.

I don't expect to document semantics of version names.
As long as there are no colons in the version name, then my
proposal should manage.  Even then, escaping colons, or using
&#xx; would work.  We already have a prototype working as a
server extension for the Netscape and MS servers.

>I think that it may be a more profitable approach to look at versioning 
>in terms of annotations and link semantics and not in terms of URLs.
>After all we may well have an MD5 URN someday [I already have a protocol
>for resolving these]. There would be little opportunity to make an
>MD5 URN work with versioning...

Annotations and link semantics might also be appropriate.  After
talking to Ari about how to present the idea, in light of his
experience with the byte-range proposal, I figured that URLs are
a good start, because the have the advantaqge that the users can
enter them directly to fetch a particular revision.

- David

Received on Tuesday, 16 April 1996 18:07:20 UTC