W3C home > Mailing lists > Public > uri@w3.org > January 2001

Re: Generic Support for New URI Schemes with Apache?

From: Al Gilman <asgilman@iamdigex.net>
Date: Mon, 22 Jan 2001 14:00:42 -0500
Message-Id: <200101221857.NAA323941@smtp1.mail.iamworld.net>
To: Rob Cameron <cameron@cs.sfu.ca>, uri@w3.org
Cc: gis-wg@gridforum.org
At 10:16 AM 2001-01-22 -0800, Rob Cameron wrote:
>Does anyone know if the following idea has been explored?  Many
>browsers that encounter URIs with unknown schemes will (incorrectly,
>I think) treat them as relative URLs.  Often this will give the
>server that served the document containing the URI a chance to
>respond constructively to the error.   Indeed, server software
>could be written to detect this case generically and dispatch
>to appropriate remedy URLs for the scheme.
>For example, with our work on the "bibp:" scheme, we find that
>some clients are hitting our server with "URLs" like 
><http://usin.org/bibp:...>http://usin.org/bibp:....  We will be using Apache
>rules to provide an appropriate response (soon).   But to deploy
>this error handling on other servers would require these
>rules to be manually configured in each case.
>If Apache had a generic support facility for unknown schemes,
>the deployment of new schemes could be simplified considerably.
>My questions for the URI WG are this:
>(1)  Does anyone know of work in this area?
>(2)  If not, would work in this area be welcome?
>(3)  If so, what issues need to be addressed?

Are you conversant with the existing work targeted to development of
service-locator services?

That is what the error recovery here might gateway into.

The recovery rules aren't necessarily site specific.  It looks as though what
the webserver needs to do is

a) on 404
b) do head matching to strip the longest BASE that actually matches your
current offerings in http: namespace
c) check the resulting tail for URI compliance
d) if in URI syntax, pass to some 'whazzat?' server on the net.
e) if success, pass results to consumer [as 300?] in lieu of the 404 you would
have sent.
f) if no match, send 404 as before

And yes, I bet that this recovery trap door into the service location service
would get lots of use.


>Robert D. Cameron
>Professor of Computing Science
>Associate Dean of Appliend Sciences
>Simon Fraser University
Received on Monday, 22 January 2001 13:51:14 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:25:02 UTC