Re: HTCP versus ICP

On Wed, 1 Jul 1998, George Michaelson wrote:

> Can you make a first-order guess at the size of the false-hit levels? small
> is very relative to the eye of the beholder.

NLANR caches showed about 5% false hits for cache digests the last time I
could check. I have no information on false hits for ICP.

Also note that the penalty for a false hit with cache digest is sending an
HTTP request and receiving a 504 reply (presumably using a persistent
connection to the peer). This penalty will not be "large" in most cases. I
bet it would be comparable to a delay introduced by ICP-like protocols.

Alex.
Received on Tue Jun 30 1998 - 23:06:27 MDT