Re: HTTP1.0/HTTP1.1, persistent connections

On Mon, 25 Sep 2000, Henrik Nordstrom wrote:

]> I am asking you this, because this way would be extremely benificial for
]> an internet link by satellite, since a round trip time is over 0.5
]> second. Therefore establishing a TCP-connection alone takes over 1.5
]> seconds at least.

If the *Round*TT* is 0.5 seconds, theoretically the active open side has
the connection established after 0.5 seconds, and the passively opened
side after 0.75 seconds. Double the values, if the one-way trip time is
0.5 seconds - arriving at what you said.

]Another viable approach for high latency links is to have a pool of idle
]TCP connections already established.

Ok, possessing an already opened connection still saves on the overhead,
and is probably beneficial for the satellite link. Nevertheless, here are
my EUR 0.02 on the issue: Modern implementations keep a cache of RTT,
ssthresh, etc. in the routing table. If a connection is opened to the same
destination within the expiry period, the remembered values ought to be
used for a more optimal start-up phase. Of course, this does not shorten
the establishment in any way. Also, if an opened connection is idle for
some time, the slow start after idle jams in. If that happens, you only
save the connection establishment overhead as compared to establishing a
fresh connection.

Le deagh dhùrachd,
Dipl.-Ing. Jens-S. Vöckler (voeckler@rvs.uni-hannover.de)
Institute for Computer Networks and Distributed Systems
University of Hanover, Germany; +49 511 762 4726

--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html