- From: Scott Lawrence <lawrence@agranat.com>
- Date: Thu, 30 Nov 2000 09:16:47 -0500
- To: http-wg@cuckoo.hpl.hp.com
- CC: dillon@hns.com, Patrik Carlsson <patrik.x.carlsson@era.ericsson.se>
dillon@hns.com wrote: > > > You have an HTTP response header (around 250 bytes) which goes > in its own segment and then all succeeding data is sent in TCP segments as a > stream of bytes. > No overhead other than the TCP/IP overhead. > > The latest standard (HTTP 1.1) has provisions for compression and "chunked" > transfers > which change this, but I haven't seen these used in any real-world situations > yet. Chunked encoding is frequently used in situations where the server might not know the size of the response body at the time it is generating headers. This is often the case for CGI and other dynamic interfaces. If chunked encoding is used, there is a short chunk header that preceeds each chunk - no more than 10 bytes; the chunk size is up to the server, but is typically either the record size of the buffer it is using to read the dynamic content, or the buffer size of its network buffers (depending on which memory it is optimizing for). -- Scott Lawrence Architect lawrence@agranat.com Virata Embedded Web Technology http://www.emweb.com/
Received on Thursday, 30 November 2000 06:23:18 UTC