- From: <jg@zorch.w3.org>
- Date: Thu, 26 Sep 96 10:36:12 -0400
- To: John Franks <john@math.nwu.edu>
- Cc: Benjamin Franz <snowhare@netimages.com>, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
I agree with the sentiment that HTTP/1.X is a human readable protocol, and should be kept so in its current incarnation.... We'll revisit this whole problem with a binary encoding someday, and the sins of HTTP are already massive in this area. However, if there are two equally good names for a header, and one is shorter, pick the shorter. And think a bit to see if a good short name occurs to you. Similarly for the encoding of values. Remember that on a 14.4 modem, a character costs you of order a half a millisecond; the milliseconds add up quickly at that rate (and people use the Web at even slower speeds; cellular telephone technology makes that look fast). Human perception is in the 30 millisecond range, and people start thinking things are slow when you get into 10ths of seconds. The bulk of web usage today is dialup, at 14.4 or up to 28.8K baud, not the speeds most of us use in the office. - Jim
Received on Thursday, 26 September 1996 07:51:36 UTC