Re: Caching of Style sheets
Please let me know if my edit marks (denoted with ) are incorrect:
On Sun, 19 May 1996, Clark Breyman wrote:
> While my suggestion might be restated "Why don't we put a UUID in every
> web content object?", I thought it best to limit my suggestion to CSS
> as there is still the opportunity to influence the initial development
> of CSS enabled UA's.
> I agree that UA[s] could cache style sheets as the[y] cache HTML documents,
> however URLs are not adequate as keys for caching.
This is www-style, not www-proxy or www-caching. This is a generic situation
with all objects on the web, and is something that other arenas address with
much more rigor. While it may be tempting to try and solve larger
problems within a smaller context, it's a rathole. We've been seeing
this in the VRML community too.
> They indicate nothing
> about the version of the content identified, whether it was copied for
> somewhere else, etc.
There are facilities in HTTP to communicate this "meta-information",
information about the object. Whether the UA decides to use it is
> Think about the whole site replication issue -
> something a URL based caching scheme could [not?] deal with, but a UUID caching
> scheme could.
This is not going to be a big problem in practice, since clients will
typically only interact with one known replicated "site" in such a
situation. Furthermore, without a rigorous certification system, how do
you prevent spoofing? If I make (or select on accident) my private UUID
to be the same as Netscape's UUID, will my style sheet affect your
subsequent visits to Netscape's pages? That could get interesting fast.
> In addition, my guess is that style sheets will become
> reusable licensable elements like Java applets. A UUID-keyed caching
> scheme would allow an author to publish a style sheet, multiple sites to
> use it and the UA to recogize that reuse.
Similarly, multiple sites can use the same URL, pointing to the source
of the style sheet. They give up change control, but they can override
anything in the first with a second style sheet. But I'm presuming
change control would still be an issue in a "UUID" caching stream anyways.
It sounds like URL's would be fine to you if only they were persistant,
described revisions, and had a layer of indirection during resolution.
What you want is the windmill of URN's, which turned many of us into Don
Quixotes for a good stretch of time. Check out
I think specifying style sheets using a URI (as it is) is fine.
firstname.lastname@example.org | We're hiring! http://www.organic.com/Home/Info/Jobs/