RE: Resource Gloss (Human Readable)

[Michael Brennan <Michael_Brennan@Allegis.com>]

> From: John Cowan [mailto:jcowan@reutershealth.com]
> [...]
> The purpose of catalogs is to prevent your web server from being
> hammered by millions of sites all trying to download your DTD to
> validate your widely distributed documents.
> 
> With catalog technology widely disseminated, this is not necessary.
> Each site can make a local copy of the DTD, and then use a 
> local catalog
> to map the central, standardized URI to a local URI or file name.

That's what I thought, and that certainly makes sense. This seems somewhat
orthogonal to RDDL to me. RDDL could be used, though, as a sort of bootstrap
for a catalog. I could see, for instance, a catalog mechanism that first
tries to locate resources associated with a namespace URI in its local
catalog. If it can't locate what's being requested, the catalog checks to to
see if the the namespace URI is a resolvable URL. If so, it does a "lucky
dip" to see if RDDL is at the end of the URL. If so, the catalog retrieves
the associated resources (or at least those of interest), and then places
them in the local catalog for future use.

This makes me think, now, that perhaps RDDL should include an "expires"
attribute for resources to indicate to such a catalog that it should refresh
local copies of resources after a period of time? (One could perhaps just
use the appropriate HTTP header for this, so long as the resource is fetched
via HTTP. However, in practice, this header seems to be rarely used.)