> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On
> Behalf Of Sean M. Burke
...
> At 10:36 2002-03-14 -0800, Nick Arnett wrote:
> >[with Python] I'm not seeing long, mysterious time-outs as I
> occasionally
> >did with LWP.
>
> I have never run into this problem, but I have a dim memory that
> you may be
> alluding to what is a known bug not with LWP, but with old versions (long
> since fixed in modern Perls and/or CPAN) of the socket libraries in IO::*.
I'm very diligent about updating, so I doubt if I was seeing an old bug.
What I would see would be a series of time-outs, usually no more than 10 in
a row (I limited re-tries to 10, with increasing delays between them in case
it was a server busy issue). But I should make it clear that the bug
producing the error message out of expat was completely separate.
> >Following an LWP request through the debugger is a long and convoluted
> >journey...
>
> Are you referring to perl -d, or LWP::Debug?
Sorry for not specifying. I was using the ActiveState graphical debugger on
Windows, although sometimes the code was actually running on Linux. Same
behavior on both, though. I did give LWP::Debug a shot, but still could see
where the error code was getting introduced. Wish I could recall better
specifics, but it's been a few weeks. As I recall, the server was returning
an error, suggesting that there was something malformed about the request I
sent it, and that error was being mistranslated in the expat DLL... and I
recall having trouble even figuring out where expat got involved in the
mess.
> Maybe I should write an addendum to "lwpcook.pod" on figuring out what's
> going wrong, when something does go wrong. The current lwpcook really
> needs an overhaul, and once my book /Perl & LWP/ is done (hopefully it'll
> be in press within a few weeks), I hope to send up some big doc
> patches to
> LWP, at the very least revamping lwpcook and then going into each
> class and
> noting in the docs whether a typical user needs to bother knowing about
> it. (E.g., you need to know about HTTP::Response; you do /not/ need to
> know about LWP::Protocol.)
That would really be good. Examples, examples, examples. I learn by doing,
not by reading, and I think there are a fair number of people like me out
there.
> In short, if people want to see improvements to LWP, email me and
> say what
> you want done, and I'll either try my hand at implementing it, or
> I'll pass
> it on to someone more capable.
A re-try mechanism would be terrific. Mine is fairly straightforward. The
parameters are a max number of tries, a delay factor that optionally rises
with each try, and a logging method that details as much as possible about
each failure. The latter is where some work on the internals would be
helpful, to disambiguate error messages as much as possible. Perhaps a
simple way to kick in LWP::Debug with appropriate parameters and log the
results if repeated failures occur? I always want to see exactly what the
outgoing request was and the server's actual response, so I know whether the
request is munged or the server is being difficult... not that that's always
clear.
Nick
--
This message was sent by the Internet robots and spiders discussion list
([EMAIL PROTECTED]). For list server commands, send "help" in the body of a message
to "[EMAIL PROTECTED]".
