- From: Hallam-Baker <hallam@ai.mit.edu>
- Date: Mon, 30 Jun 1997 10:39:51 -0400 (EDT)
- To: delabeau@iniki.gsfc.nasa.gov
- Cc: GK@acm.org, http-wg@cuckoo.hpl.hp.com
> Either a prefix like /cgi-bin or suffix like .cgi may be relevant, depending > on the server config. In the context of 1.1 compliant servers cgi or otherwise plug in enabled servers should by default be explicitly stating that the content is not cacheable. The heuristics whereby 1.0 proxies guessed what material was static and what was dynamic are legacy issues that we should not be carrying over. There can be no assigned paths in HTTP that will not cause problems. Consider a machine where the file system is served via HTTP rather than NFS. In this context it would make perfect sense to access a file /foobar.com/cgi-bin/a.cgi as a static file. "Assigned paths" sounds pretty much like what we used to call a "land mine". Some arbitrary restriction embedded in a protocol that means that a whole class of problems have to be considered a special case. If people want a robot control protocol we should write the damn thing into the spec. Its quite an easy problem to solve because most of the Web crawling is done by the index compabies who are probably likely to play ball. Phill
Received on Monday, 30 June 1997 07:41:25 UTC