[EMAIL PROTECTED] wrote:
>
> On 8 Aug 2000, at 18:25, [EMAIL PROTECTED] wrote:
>
> > I tried all sorts of timeouts and that never helped at all. I even threw in a
> > ALARM which didn't have any affect unless I used a well known URL. I have
> > seen LWP waiting for over a hour on the URL. Around lunch time it did respond
> > back after 5 mins with a error that it could not connect with a
> > $ua->timeout(10). A later attempt it just hanged for over 30 mins till I
> > killed it. Seems to me a lot of the timeout problems could perhaps be solved
> > if one could understand what is happening with these various URLS and their
> > server responses to make socket/select more responsive for a wider variety of
> > URL's encountered since this is what LWP relies on for the connection.
> >
>
> If you don't mind, can you provide a few example urls? It sounds
> like we're running up against similar problems, except that our
> system is running the requisite 3 minutes before failing (so far).
> Unfortunately 3 minutes is still too long and I would like to reduce the
> timeout.
>
> Does anyone know if the timeout feature is being worked on?
>
> Regards
> Bob Minnick, Northern Webs
>
Hi Bob,
I think we 'met' back when I had my fishutah.com/huntutah.com sites; could
have been '95/'96 or so; small world.
My belief is that no one is working these problems. I've had them
too, almost solely with the parallel calls. I had to put a few
timers in there, and have my (LWP Parallel) callback routine send
different return codes (that are clearly documented in the parallel LWP
docs) causing a shut down of that or all open channels (depending on
the callback routine's judgement).
One problem I've seen with that solution is when you get in a position
of waiting on only *one* connection and that one connection is not
going to respond. In that situation your callback routine doesn't get
called (no data recieved, so why call it?) so it can't shut LWP down.
In most cases you're getting valid data from other sites while the
dead/slow site aren't responding, so your callback routine has the
ability to do it's (shutdown of LWP) stuff. One way I've been able
to encourage this better behavior is to set my internal timers such
that I always shut down some data from arriving - say when about
90% of the expected data has been recieved.
Said another way, after I have ~ 95% of my data, the remaining sites
generally don't respond, and that puts me in the bad position of
my callback routine not being called, exactly when it should be
shutting LWP down.
Good luck. It's fun when its not incredibly frustrating. :-)
Steve
--
Steve Rasmussen <[EMAIL PROTECTED]>
President
Secure Solutions Inc. http://4secureorders.com
"We are what we repeatedly do. Excellence, then,
is not an act, but a habit."
-- Aristotle
