> Jason van Zyl wrote:
> > What if the log4j developers made a really R
> > tiny core package, say 10-15k that provided
> > the bare minimum for logging? Would people be
> > willing to use it then?
>
> That's exactly what I was thinking. It seems like we could get away with
an
> extremely simple implementation of the log4j Category class that just
dumps
> to stdout, or conditionally dumps to stdout based upon the value of some
> runtime property, or even one that just quitely ignores all the logging
> calls. The whole thing could probably be done in a couple of KBs (not
even
> 10-15), which would suffice for those who don't care much for/about
logging.
>
>
> Re. the interceptor point, I'm not sure that does what we need/want here.
> Remember that the problem at hand isn't messaging, it's debugging. Most
> (all) of the log output in httpclient right now is there to
> trace/diagnose/understand the inner workings of httpclient. Sending
> "events" to listeners isn't going to offer the level of detail desirable
for
> debugging httpclient.
There are two types of "fine debugging" that you would want to do :
a) debug needed to test interoperability with another web server
b) debug needed when fixing a bug in HTTP client
a) can be done with the current interceptors and the current set of hooks.
b) can't, but IMO that kind of thing has nothing to do in the CVS in the
first place. Usually, to do that, you can use your favorite IDE's debugger,
or a lot of Sys.outs (I chose the Sys.outs).
Remy
