- From: Ingrid Melve <ingrid.melve@uninett.no>
- Date: Wed, 17 Mar 1999 18:21:34 +0100 (MET)
- To: Harald Alvestrand <Harald@Alvestrand.no>, Koen Holtman <Koen.Holtman@cern.ch>, Larry Masinter <masinter@parc.xerox.com>
- Cc: Keith Moore <moore@cs.utk.edu>, web@apps.ietf.org, discuss@apps.ietf.org
At 09:26 16.03.99 +0100, Harald Alvestrand wrote: >At one time at least, the ICP protocol of SQUID would ship small documents >in UDP datagrams on the ICP response. Yes, we did. >That community is probably able to give practical, recent experience. No, it is not a good idea. >(They may still be doing it....) Some do, but it is still not a good idea - UDP fragementation - access control trouble (lack of HTTP headers) - pinelining requests give better performance ( we used to do this in the days of HTTP/1.0 ) This is off the top of my head, there is more negative experience that I have suppressed (we did do WAN UDP-packets for quite a while in the COM-MESH experiment). And of course we are changing to CacheDigest these days... Ingrid
Received on Wednesday, 17 March 1999 12:22:23 UTC