- From: Jeffrey Mogul <mogul@pa.dec.com>
- Date: Fri, 19 Jan 96 12:03:34 PST
- To: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
The persistent connections subgroup has made a lot of progress. We held a meeting last Friday, at which we came to a consensus on what amounts to a minor variation of the experimental Keep-Alive: and Connection: headers. The primary issue was how to deal with HTTP/1.0 proxies in a reliable way, and we believe we have a simple solution. There is still some debate about whether the spec should include a "Proxy-connection:" header, which has been used by a few vendors without any prior documentation, and which is clearly prone to failure in certain contexts. We have several remaining issues. One is how responses should be terminated. The two possible options (besides Content-length:, which is not always an option) are chunking and multipart; we are apparently agreed that one of these ought to be available in all cases where a client has requested a persistent connection, so that the servers will have something to depend on, but we are not agreed which option(s) should be required by the spec. We are also expecting to consider the issue of "sticky headers" (previously these had gone under a different name), which allows a client or server to avoid retransmitting headers that don't change between messages on a single connection. -Jeff
Received on Friday, 19 January 1996 12:15:17 UTC