Jon Anderson wrote:
...
SetEnvIf User-Agent ".*MSIE.*" nokeepalive ...
...
I seem to recall this being due to a bug in Internet Explorer that keeps
the connection open longer than necessary, tying up server resources.
George Schlossnagle has a formula for calculating the win/loss of using
Richard Lynch wrote:
I *think* both client and server have to agree/support keep-alive, and
both have to be ready for the other guy to close the connection even
though they agreed in advance to keep-alive.
So you can use keep-alive, and it can work nifty, but ya gotta be
ready for the server/cli
I *think* both client and server have to agree/support keep-alive, and
both have to be ready for the other guy to close the connection even
though they agreed in advance to keep-alive.
So you can use keep-alive, and it can work nifty, but ya gotta be
ready for the server/client to claim to support
httpd.conf search for Keep-alive
Some old old old browsers will puke on it, though, I think...
We're talking like Netscape 2.0 though, as I recall...
I got swiss-cheese memory, so better double-check that. :-)
On Tue, March 21, 2006 5:16 pm, Jon Anderson wrote:
> I've been looking at a proble
Chris Shiflett wrote:
In that case, I think a good next step would be to examine the HTTP
request. One guess is that the request you type in manually with
telnet is HTTP/1.1, and the one being sent in your SOAP request is
HTTP/1.0.
You are actually quite correct... The requests are being sent f
Chris Shiflett wrote:
Chris wrote:
If you're doing an exit() or die() or the script stops executing
that's like you closing the connection - so apache is going to
close the connection (as it should).
The connection Jon is talking about is the TCP connection, just in case
that's not clear. I
Chris wrote:
If you're doing an exit() or die() or the script stops executing
that's like you closing the connection - so apache is going to
close the connection (as it should).
The connection Jon is talking about is the TCP connection, just in case
that's not clear. I'm not sure what connecti
Jon Anderson wrote:
Keep alives are definitely configured in the server - I can
request scripts multiple times manually from a telnet client.
In that case, I think a good next step would be to examine the HTTP
request. One guess is that the request you type in manually with telnet
is HTTP/1.1
Chris Shiflett wrote:
Check the KeepAlive configuration directive in httpd.conf.
Keep alives are definitely configured in the server - I can request
scripts multiple times manually from a telnet client. It seems to be
just SOAP requests in PHP that get mucked up - even if I send a
"Connection:
Jon Anderson wrote:
Chris wrote:
That's the design of http - it's stateless. Each connection is treated
separately and as such closes itself when it's finished.
If you want to keep it open you'll need to create your own "service"
that listens to a port and responds accordingly - which can be
Chris wrote:
That's the design of http - it's stateless. Each connection is treated
separately and as such closes itself when it's finished.
If you want to keep it open you'll need to create your own "service"
that listens to a port and responds accordingly - which can be quite a
lot of work.
Jon Anderson wrote:
It seems to work okay, except that at the TCP level, the client
keeps initiating new connections for every soap request rather
than using a single connection for multiple requests. One
possible reason for this is that the server sends a "Connection:
close" HTTP header after th
Jon Anderson wrote:
I've been looking at a problem with our software. It uses Flash with
SOAP to connect to a PHP backend. It seems to work okay, except that at
the TCP level, the client keeps initiating new connections for every
soap request rather than using a single connection for multiple
I've been looking at a problem with our software. It uses Flash with
SOAP to connect to a PHP backend. It seems to work okay, except that at
the TCP level, the client keeps initiating new connections for every
soap request rather than using a single connection for multiple
requests. One possibl
14 matches
Mail list logo
