- From: Peter Ball <pball@hab-software.de>
- Date: Tue, 03 Sep 1996 12:21:20 -0700
- To: www-talk@www0.cern.ch
As written in http://www.w3.org/pub/WWW/Protocols/HTTP/HTRQ_Headers.html, currently only one pragma is defined: no-cache. I believe that it would be useful if an extension would be made to this. Currently the no-cache pragma instructs proxies and local caches not to cache the document, however there are situations where it still makes sense to cache a document locally (in the browser) but not on the internet (in a proxy). A "no-proxy" pragma could be defined for this purpose. If some CGI program produces some client dependant output (e.g. optimized for the browser type (by checking the HTTP_USER_AGENT enviroment variable) or for the user (perhaps by checking the user's address in the REMOTE_HOST enviroment variable)), then this information could be cached on the client (for the client it is always the same) but not on the internet since the document isn't always the same. Consider the following: the client requests an address which specifies a CGI program with some parameter that tells the CGI what data to return. The CGI retrieves the data from a database and writes some HTML which is customized for the client (lets say the country code in REMOTE_HOST is checked to see what language to use). The user will always get the same thing so it makes sense to cache this locally. But on the internet there are probably documents (under the same URL) that are floating around that are slightly different and should not be cached. In this case a "Pragma: no-proxy" could be a nice addition for any future HTTP specification. On the other hand maybe I'm talking a load of rubish.
Received on Tuesday, 3 September 1996 06:22:33 UTC