PURLs
weibel at oclc.org
weibel at oclc.org
Wed Jan 10 08:18:37 EST 1996
Marc Salomon writes:
> persistence is but one slice of the larger (largest?) problem.
Marc is exactly right. PURLs, by providing an additional level of
indirection, are intended to be only *one* component of an incremental
solution to the larger problem. The larger problem requires the
evolution of services that librarians will recognize as cataloging
systems applied to networked resources.
> replication is equally as important for certain documents. one problem
> i have w/purls was that it was a one-to-one mapping of purl->url, and
> for many applications this hinders scalability. why can't a purl server
> respond with multiple uri: fields pointing to replicated documents and
> the client can decide which to retrieve?
Marc's replication issue is important, its just not the problem that
PURLs are intended to solve. Our view is that even with a complete
URN solution, resolution should return *exactly* one URL.
Choosing a URL among several alternatives can be made according to
political, technical, geographical, and economic factors. Might not
servers be a better place than browsers to implement such decisions?
Besides, keeping a marketplace of browsers in sync (indeed, changing
them at all) is not a happy prospect.
Among the virtues of PURLs is that they work now, with no modifications
to browsers. tools to manage the replication problem can be built on
top of PURLs.
> can a client/proxy cache the location: <url> as a substitute for the
> original one or will you use the moved temporarily http response code,
> and if the latter doesn't this present scaling problems as you double
> the number of transactions required to fetch each request?
The PURL server returns a URL to the client, which the client fetches
in a conventional manner, and caches normally.
> but i do think that this is a good first step towards experimenting with a
> level of indirection in web document access.
Thanks! Stop in at http://purl.oclc.org and get a PURL for your homepage.
More information about the Web4lib
mailing list