SRS will soon stop working :-)

Claude Scarpelli claude at lovelace.infobiogen.fr
Tue Oct 1 06:13:06 EST 1996


We have noticed that most (all ?) SRS WWW sites around the world are
configured such in a way that they are not compatible with the
emergence of WWW cache.

When the browser's client is configured to use a WWW cache, the SRS
queries issued by the user may be processed with another client
parameters.

Here is an example:

an SRS session usually begin with (BEN site is used as an example):

GET http://ben.vub.ac.be/srs/srsc

which retrieve the main SRS page. On the server side, the wgetz
program has been run. To keep the session context, it creates a
temporary file whose name contains the time and the IP address of the
client. This filename is propagated to the form sent to the user as a
hidden field.

Therefore it is critical that a different file is created for each
different user. Unfortunately, this may not be the case a WWW cache is
used. The former URL does not contain any magic information (neither
explicit) to tell the cache to NOT cache the document. So there is a
strong probability (we have easily reproduced it) that the document
received (http://ben.vub.ac.be/srs/srsc) by the user has been cached
by the WWW cache, resulting in:

1) the document (the hidden field used to retrieve the context)
   contains context information relevant to another user (actually the
   one for which a cache miss occurred)

2) wgetz has not been run on the server, so no context for this
   particular user has been created.



Therefore, in case the SRS WWW sites remain configured as they are
currently, users are going to face more and more difficulties to
access SRS services.

How can it be fixed ?

1) configure the WWW cache to NOT cache URL containing srs/srsc. Not
   practicable, since user may not have enough authority to modify
   the configuration of the cache they are using.

2) Configure the SRS site in such a way that the magic strings
   /cgi-bin/ appear in each URL. WWW cache never caches a document
   when the URL contains this magic string.

3) Have wgetz send a protocol element (Pragma: no-cache) with each
   documents (especially the very first one). Cache will honor this
   information by not caching the document. Of course, it requires
   modification on the wgetz source code. We will ask the author to
   proceed this way, since it is the best way to communicate with a
   WWW cache


At our site we could only apply the second solution, with the help of
the SRS documentation which gives you all the information to modify
the SRS WWW URL.

This is why the SRS WWW server at INFOBIOGEN can be reached at the
following URL:

http://www.infobiogen.fr/srs/cgi-bin/srsc

We strongly encourage the SRS source maintainer to implement the
solution 3 discussed above, at least in the next release. In the
maeantime, we urge every SRS WWW site to consider modifying their
configuration.



Claude Scarpelli     -- Jean-Marc Plaza     -- INFOBIOGEN
claude at infobiogen.fr    plaza at infobiogen.fr    bioinfo at infobiogen.fr


-- 
------------------------------------------------------------------------------
Claude Scarpelli                        | Defenestrate: to exit a window
INFOBIOGEN ::= INFOrmatique appliquée à | onscreen. (Time International
l'étude des BIOmolécules et des GÉNomes	| Vol 146, No. 20, Nov 13, 1995)




More information about the Bio-srs mailing list