Re: Comments on HTTP draft [of 23 Nov 1994]
Roy T. Fielding (fielding@avron.ICS.UCI.EDU)
Fri, 02 Dec 1994 13:08:47 -0800
> Looking at IMS on a packet by packet basis, in the presence of possibly
> large headers greater than 512 bytes, having the client cancel will save
> 2 packets in the best case, and be no worse in the worst case. I sent
> a message on tis to www-talk during chicago with more analysis. my hands
> aren't up
> to recreating it at the moment, and I think the message dissapeared
> in transit. i'll see if i can find it again.
Good luck. Given that a typical IMS request message is
64 bytes + length(URL) + user-agent (if any)
and a 304 Not Modified response is typically 98 bytes, I haven't got a
clue as to how you could come up with such figures.
That means (aside from the connection packets) one packet for the request
and one for the response. In contrast, a normal request will generally
buffer 16 packets before the client even has a chance to read the response --
cancelling at that point is useless.
......Roy Fielding ICS Grad Student, University of California, Irvine USA
<fielding@ics.uci.edu>
<URL:http://www.ics.uci.edu/dir/grad/Software/fielding>