Paying for someone else's traffic?

Hmm .. just thinking aloud here. "miss_access" is there to ensure that
neighbour proxies can't use your proxy as a parent without your
permission, right? The idea is, if some other proxy requests an object
that isn't already in your cache then your cache *won't* retrieve the
object for them (because you could be paying the bill for their Web
surfing).

What about refreshes?

The log entry that triggered this looks something like this:

888631512.070 779 a.b.c.d TCP_REFRESH_HIT/304 103
  GET http://www.sunday-times.co.uk/standing/tim/wkmbaroff.gif -
  DIRECT/www.sunday-times.co.uk -

IP address "a.b.c.d" is in some Class B address that is outside our own
Class B address - so a miss should be denied. But this isn't a miss;
clearly our proxy has responded to an ICP request by saying "Yes, we have
that object" but the HTTP request has been for a *refresh*, so our proxy
has gone off to the remote site. Now, "304" is "not modified", so our
proxy hasn't retrieved the object again and just passed back that which is
in our cache.

Right?

Okay, what about this pair:

888633275.342 1 a.b.c.d UDP_HIT/000 64 ICP_QUERY
  http://freebeer.com/MidsHotStuff/cblogo.gif - NONE/- -
888633276.244 891 a.b.c.d TCP_REFRESH_HIT/200 7123
  GET http://freebeer.com/MidsHotStuff/cblogo.gif -
  DIRECT/freebeer.com image/gif

As before, our proxy responds to an ICP query but the subsequent HTTP
query from the remote proxy is for a *refresh*. This time the code is
"200" and the file size is 6-7Kb (instead of a few hundred bytes).

I *am* reading this correctly - I *am* paying for someone else's Web
browsing traffic, right? :-/

Can/should "refresh" be made equivalent to "miss" for the purposes of the
"miss_access" keyword?

All comments/input appreciated.

dave
Received on Fri Feb 27 1998 - 19:01:29 MST