- From: Henrik Frystyk Nielsen <frystyk@w3.org>
- Date: Sun, 30 May 1999 12:38:39 -0400
- To: danielh@econ.ag.gov, www-talk@w3.org
At 18:54 29/05/1999 -0400, danielh@econ.ag.gov wrote: >Given that one might expect a major use of a maintained connection to be >retrieval of inlined images, and since the user-agent has to parse the >html document before knowing what images to request, it seems that the >"pipeline" method is less useful (although a second connection, with >multiple requests for various images could be used..) Note that the client doesn't have to parse the *whole* document before issuing new requests - often there are enough links embedded in the first chunks of HTML to enable the client to submit new pipelined requests almost immediately. You can see a more detailed explanation of the potential benefits of pipelining in our sigcomm paper: http://www.w3.org/Protocols/Performance/Pipeline.html Paul Barford also has a very interesting paper on pipelining but I don't have a link handy. >>If a timeout system is used should the server send a response and >>Connection: close header before it closes the connection? >>Even though there might not be a request to respond to? >That strikes me as unnecessary. a maintain-connection capable browser >should be smart enough to know when to give up, but may not be smart >enough to know what to do with a unsolicited response. No unsolicited response - the client is interested in dropping the connection when it is done because having the server closing the connection may force the client to recover a pipeline. The server can always drop the idle connection when low on resources. Henrik -- Henrik Frystyk Nielsen, World Wide Web Consortium http://www.w3.org/People/Frystyk
Received on Monday, 31 May 1999 13:12:46 UTC