- From: <touch@isi.edu>
- Date: Sun, 30 Jun 1996 13:05:31 -0700
- To: Harald.T.Alvestrand@uninett.no, touch@isi.edu
- Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
> the analysis you posted seems to be based on the assumption that the > 'Net has room for the transactions, and that RTT*bandwidth is the limiting > factor. Yes. > What I at UNINETT think is my problem that p-http will solve is the behaviour > of "n" simultaneous HTTP clients and a connection (the US link) that has > packet loss due to congestion, where I think the persistent HTTP sessions will > back off more gracefully than the per-transaction ones, > leading to fewer dropped packets and better overall throughput. > Is there any analytical result for this case? Not yet. However, if p-HTTP backs off better than the current case, clients may opt for the latter. "Conventional wisdom" is that, when provided with various levels of priority that can be controlled by the user, the user will always chose the highest priority. Joe ---------------------------------------------------------------------- Joe Touch - touch@isi.edu http://www.isi.edu/~touch/ ISI / Project Leader, ATOMIC-2, LSAM http://www.isi.edu/atomic2/ USC / Research Assistant Prof. http://www.isi.edu/lsam/
Received on Sunday, 30 June 1996 13:19:27 UTC