httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alexei Kosut <>
Subject Re: Problems with Apache 1.2b1 and HTTP/1.1 pipelining (fwd)
Date Thu, 19 Dec 1996 20:55:24 GMT
On Thu, 19 Dec 1996, Chuck Murcko wrote:

> Henrik's observations look good. It looks like we're using the Keepalive
> value to terminate pipelined requests, also. That's why Henrik reports
> Apache closing the connect after 5 responses, I suspect. I agree that
> packing short responses before flushing buffer is a Good Idea.

Of course we're using keepalive values for pipleined requests -
they're the same thing. The only thing about pipelining is that it
requires the server to queue requests (wich Apache does).

The historical reason for the KeepAlive number at all is that the
original version of the Apache keepalive code used the same pool for
all the requests, which meant that unless you limited the number of
requests per connection, the member requirements got prohbitive.

This has been fixed, and really the restriction is somewhat arbitrary
and unneccessary. Ideally, you'd just have KeepAlive On/Off, and it
would accept unlimited requests per connection. I can't think of any
problems with this, in the way of overhead or memory or server
overloading (presumably, the client would just open up another
connection and pump in 5 more requests - the timeout value would be
unaffected). In fact, one might even consider the current behavior a
"bug". (though it'd be a stretch).

> I'm not sure what might be causing the lost bytes, though.

Yeah, nor do I. This is worrisome.

Alexei Kosut <>      The Apache HTTP Server

View raw message