- From: David W. Morris <dwm@xpasc.com>
- Date: Fri, 20 Mar 1998 10:26:13 -0800 (PST)
- To: Michael Wexler <mwexler@nny.com>
- Cc: http-wg@cuckoo.hpl.hp.com, Andy Norman <ange@hplb.hpl.hp.com>
Have you reviewed RFC 2227 -- "Simple Hit-Metering and Usage-Limiting for HTTP"? Seems like this RFC is trying to solve your problem. Dave Morris On Thu, 19 Mar 1998, Michael Wexler wrote: > As a web measurement analyst, one of my worst problems is dealing with > caching. Why do we use the code 200 for everything? Why not design the > spec in a graduated fashion: > > If browser has url in its local cache, it still sends a get request to the > server, but an option says "I already have it, just letting you log the > request". The success code is a 209, "user has non-expired data in cache". > > If a non local cache has the data, same system. The option can be > different, and we can even use a different code (210), but for the most > part, we should just let the 209 mean "cached request". > > This solves many problems: > 1) path analysis of a user's visit > 2) advertising requests (not perfectly, given the IAB's recent standards, > but better than nothing for smaller sites) > 3) pages per visit calculations are accurate > 4) this is a minor increase in bandwidth compared to not sending the > request at all, and is far superior to eliminating caching.
Received on Friday, 20 March 1998 10:30:19 UTC