- From: <Harald.T.Alvestrand@uninett.no>
- Date: Mon, 01 Jul 1996 10:41:14 +0200
- To: touch@isi.edu
- Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
(duck)
> touch@ISI.EDU:
> PS - a recent Internet Draft discussed the issue of TCP connection
> control, and aggregation across bottleneck links (although not the
> primary issue of the draft):
> draft-touch-tcp-interdep-00.txt
thanks - I had that document in my cache to read.
>
> The general conclusion we reached was that there is a good argument
> for controlling some of the parameters of multiple TCP connections
> to account for aggregate behavior, such as the graceful behavior
> over bottlenecks you describe.
>
> This is the case not only for multiple TCPs from a single host, but
> also for multiple TCPs from a set of hosts with a similar path.
This seems like a good thing to research; at the moment, I wouldn't
even try to guess which destinations are "similar path" and which
destinations are special cases because of remote-system problems.
> This is a problem that should be addressed anyway; once addressed,
> the need for persistent HTTP weakens even further.
If we could replace kernels at the same speed we replace browsers and
servers, yes, perhaps....in many ways I see p-HTTP as a reaction to the
fact that the Net with TCP is NOT a good environment for transaction-like
behaviour; what we can do to change that underlying fact, we should..
Looking forward to seeing further work in this area!
Harald A
Received on Monday, 1 July 1996 01:48:44 UTC