I got this mail address from the Apache LOG4J web-site and I understand
this is a forum where I can express some input regards the logging
mechanisms in the standard JDK.
We have a large investment in LOG4J, it is the standard API used in all
of our java projects and we have been increasing users of it for over 1
year. It is simple, reliable and very configurable - key features for
us. It was chosen after an extensive review of the other available
logging packages, including a fairly complex in-house development.
Moving to anything else would be a significant task which we would be
reluctant to undertake without a compelling reason.
About the only reason that would cause us to change would be the
inclusion of a standard logging API in the JDK as we aim to be
standards conformant wherever practicable. However, the proposed JSR47
logging standard for JDK 1.4 falls far short of our expectations and
would be a major step backwards for us. I won't re-hash the issues as
these have been extensively documented on the log4j web site.
We are unlikely to adopt JSR47 for our use as it stands, and I suspect
we are typical of a large number of existing users. LOG4J has evolved
tremendously from its first incarnation due to feedback from people who
actually use it, to the point where it is now a mature product which
*works today*.
I once attended a talk by Gosling who stated that one of the
fundamental principles that was adopted in developing the Java API was
that it was better to leave a feature out than to include one that was
only half-ready, as adding features later to fill a gap was much easier
than fixing something that was broken but already in use.
Logging is probably one of the hardest areas to standardise, as almost
everyone can easily develop a simple logging API and there are probably
thousands of different home-grown APIs in use. I would urgently suggest
to SUN that they either adopt LOG4J as the standard logging package, or
pull logging out of the JDK1.4 (merlin) release entirely and work
through the concerns that clearly exist before making any decision on
standardisation.
regards
Stephen Ashcroft