Re: "Hits" pragma

>Are you suggesting that servers only have a right to see aggregate 
>statistics, and not per-request lists?  (i.e. "32 access of this url by 
>23 different hosts", not "hosta, timestampa; hostb, timestampB;"... etc?

Yes.  If you are not given the machine name, matching it with
timestamps isn't worth the bother.

>Also, how does the proxy determine "# individuals", or did you mean "# of 
>different IP addresses serviced"?

The former, using the same methods the server would.  The proxy may
know a lot more information about each user than it is willing to pass on.

>> Like Andrew mentioned, this is best done by passing a URL to the
>> origin server that tells it where it may retrieve a sanitized summary
>> of the data.
>
>I much prefer Buchard's model of sending it with the refresh request 
>rather than forcing a separate transaction for the data.  If a separate 
>request were made, the proxy not only has to make that data 
>HTTP-accessible to the outside world (a problem for proxies behind 
>firewalls) but also has to have some heuristics as to how long that data 
>is around.  With compression it should not add significantly to the size 
>of the request.

Except for the minor detail that the last action taken by a cache
is to throw away the cached document, not make a new request on it.
There is no reason to require the proxy to make a new request just
to inform the server (which in most cases doesn't care anyway) of
the last batch of saved hit info.

Note also that the URL for where to retrieve the summaries need
not be the same as the proxy sending it.  Of course, proxies behind
secure firewalls won't be sending you any data in any case.


 ....Roy T. Fielding  Department of ICS, University of California, Irvine USA
                      Visiting Scholar, MIT/LCS + World-Wide Web Consortium
                      (fielding@w3.org)                (fielding@ics.uci.edu)

Received on Monday, 14 August 1995 18:23:50 UTC