- From: Jeffrey Mogul <mogul@pa.dec.com>
- Date: Fri, 14 Nov 97 15:05:46 PST
- To: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Dave Kristol writes:
I think what you're missing is that most servers store files (er,
entities?) already compressed as, for example, .gzip files.
I understand that. What I think you're missing is that this not
the most desirable state of affairs. On-the-fly compression would
immensely useful in reducing bandwidth requirements for non-image
data (see our SIGCOMM '97 paper, and also the SIGCOMM '97 paper
by Gettys et al.) HTTP/1.0 doesn't really support this, but we've
tried to make it possible in HTTP/1.1
We want to encourage more efficient use of the Internet, not freeze
the current (and inefficient) practice.
Having said that: I realize that there may be a conflict between the
right thing to do for Ranges with on-the-fly compression, and for
Ranges with .gzip files. And maybe the spec needs to be able to make
the distinction explicit, rather than us arguing about which single
mode should be supported?
Can anyone offer an example of a server that compresses content on the
fly and returns it in that form?
I believe that Henrik et al. have prototyped this, and their experiences
were largely the inspiration for fixing the bugs in Accept-Encoding.
-Jeff
Received on Friday, 14 November 1997 15:11:56 UTC