Of course, I'm not suggesting that all the AS code be rewritten
overnight to use JUL, that would be a big task.
But IMHO it makes sense for new stuff, like what David or we are doing
in JBM to do this. I don't see any need for a big bang approach, this
could be factored in step by step without breaking anything.
Of course first we need to research the assertion that JUL delegation is
very buggy, which if true could be a showstopper.
Adrian Brock wrote:
On Wed, 2007-12-12 at 16:01 +0000, Tim Fox wrote:
> Huh? How is using JDK logging NIH syndrome? Seems the complete opposite.
>
> Surely writing your own (jboss common logging) is NIH?
Read the discussion.
The proposal is to replace a 16k jar (that uses a simple
wrapper mechanism) with the following:
1) The use of JDK logging.
2) Rewriting all the log4j appenders (or at least
those that people find most useful) using the JDK api.
3) Subclasses of the JDK logger api (including some
unspecified factory api) to redirect to log4j
4) Persuade everybody else to use JDK logging
because we've implemented (2) and (3) in some JBoss
project.
5) Get Sun to fix/backport their bugs so we don't
have to use log4j (zero jar dependency logging)
6) Hope it is as stable as our current implementation
or (as David would have it) assume it is until
proven otherwise.
"Accordingly, I respectfully dissent." :-)