OCLC PURL vs. server redirect?

[Stu Weibel]

Peter Gorman writes:

> Since OCLC's PURL software uses standard HTTP redirects, something
> just about any web server can do, I was wondering what specific
> value there might be in maintaining a separate PURL server.  One
> advantage that comes to mind is the potential to put a nice
> interface on the redirection function (as opposed to editing the
> server's config files directly). I suppose that there could also be
> an issue of the number of redirect statements a particular HTTP
> server could handle in its config file.

Peter's supposition is correct...  one of the main advantages to using
a PURL server is that it comes with a collection of utilities that
support the management of URLs - a critical aspect of assuring the
integrity and maintenance of links.


> Can anyone comment on the performance of the PURL software vs., say,
> the NCSA server?

The routines that handle PURL resolution are integrated with the NCSA
HTTPd 1.5.1 code (because it can be redistributed and is still among
the most popular HTTP servers).  This gives us the benefit of having a
full-fledged HTTP server available as well as all the time and thought
that has gone into the NCSA server.

There are many benefits to running multiple HTTP servers, such as load
balancing.  Most large organizations run multiple HTTP servers.
Running a PURL server in addition to other servers is merely a
resource allocation issue.  A PURL server (since it's a modified NCSA
server) can easily support PURL resolution and document serving.

Without statistics on the number of documents that need to be served,
the number of PURL resolutions required, and hardware being used, it
is difficult to say whether or not a PURL server should be the primary
HTTP server for a site.  At OCLC, we keep the PURL server separate
from the main OCLC HTTP server, but do allow it to serve a limited
number of documents directly.

One last point... there is an intangible benefit to the recognition of
"purl" in a URL address like http://purl.foo.edu/stuff.  If PURLs are
widely adopted, then seeing a URL with "purl" in the domain name
suggests that the link is a managed link, maintained by an
organization with a commitment to the persistence of access.  It is
the sort of thing libraries do.

The PURL team
purl at oclc.org