- From: Roy Fielding <fielding@beach.w3.org>
- Date: Thu, 27 Jul 1995 23:36:41 -0400
- To: Harald.T.Alvestrand@uninett.no
- Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
>Content-transfer-encodings generally stay with the message in store, >and are not used to find the end of a message, but to protect the content. > >The "packet" encoding's primary purpose is to locate the end of the >message without requiring scanning for special strings or knowing the >complete length before starting transmission. Yep, and the above two paragraphs are equivalent under HTTP. The packetized CTE would stay with the message (until it was removed), and its purpose is to protect the content from a premature closure of the connection (i.e., it allows the recipient to know if it got the whole thing, and it allows the sender to apply a signature to the whole thing). >Note that an almost identical mechanism was proposed for SMTP - I think >it is still one of the "draft-ietf-mailext" drafts. Yes, I am aware of <draft-ietf-mailext-smtp-binary-07.txt>. That mechanism uses a series of separate messages (each message given a 1*DIGIT length) to send the binary data. In other words, it is a stateful message sequence, which is something HTTP doesn't do. It does demonstrate that both protocols need such a thing, and conversion between the two methods would be trivial. ....Roy T. Fielding Department of ICS, University of California, Irvine USA Visiting Scholar, MIT/LCS + World-Wide Web Consortium (fielding@w3.org) (fielding@ics.uci.edu)
Received on Thursday, 27 July 1995 20:38:00 UTC