Here are latest results. 

> ls dist/lib/log4jME.jar

-rw-r--r--   23045 Aug  1 19:40 dist/lib/log4jME.jar

> jar tvf  dist/lib/log4jME.jar

     0  org/
     0  org/apache/
     0  org/apache/log4j/
   380  org/apache/log4j/Appender.class
  1776  org/apache/log4j/AppenderSkeleton.class
  4975  org/apache/log4j/Category.class
  1024  org/apache/log4j/CategoryKey.class
  3807  org/apache/log4j/FileAppender.class
     0  org/apache/log4j/helpers/
   373  org/apache/log4j/helpers/FormattingInfo.class
  1855  org/apache/log4j/helpers/LogLog.class
   630  org/apache/log4j/helpers/NullEnumeration.class
   658  org/apache/log4j/helpers/OAErrorHandler.class
  1348  org/apache/log4j/helpers/PatternConverter.class
  1001  org/apache/log4j/helpers/PatternParser$BasicPatternConverter.class
   936  org/apache/log4j/helpers/PatternParser$CategoryPatternConverter.class
   744  org/apache/log4j/helpers/PatternParser$LiteralPatternConverter.class
  4224  org/apache/log4j/helpers/PatternParser.class
   972  org/apache/log4j/helpers/QuietWriter.class
  2814  org/apache/log4j/Hierarchy.class
   663  org/apache/log4j/Layout.class
  2002  org/apache/log4j/PatternLayout.class
  1793  org/apache/log4j/Priority.class
   290  org/apache/log4j/ProvisionNode.class
     0  org/apache/log4j/spi/
  1721  org/apache/log4j/spi/LoggingEvent.class
   160  org/apache/log4j/spi/OptionHandler.class
   941  org/apache/log4j/spi/RootCategory.class
     0  META-INF/
   167  META-INF/MANIFEST.MF

At 11:50 01.08.2001 -0500, you wrote:

>> -rw-r--r--   1 544      cgu    29103 Aug  1 14:41 dist/lib/log4jME.jar 
>
>Hey, that's pretty cool.  Especially in that it's using the log4j code base 
>unchanged. 

The code base is similar but different.

>I was thinking something a bit more radical in it's impact on log4j's functionality, 
>but probably much smaller in footprint.  

What's the point of a 5K footprint?

>Although I haven't tried it, it seems like we could create a new 
>org.apache.log4j.Category class that turns things like addAppender and callAppenders 
>into no-ops, sends all of the debug, error, fatal, info and warn messaged to stdout 
>or stderr, and always returns the same singleton value for getInstance.  If we wanted 
>to be clever we might add the ability to turn logging on or off per category or 
>message type.  
>
>Would support for the Category methods alone cover most everyone's use of log4j?  If 
>so we should be able to support that in a single class or so, probably well under 5k 
>in binary, and still maintain API compatibility with log4j.
>
>Ideally I think we should lobby the log4j folks to create the minimal log4j binary, 
>which may entail a bit of refactoring to get it all to work cleanly (I'm a little 
>uncomfortable with dropping in a radically different implementation of the same class 
>name, defining a Category-esque interface would seem to be better), but I think 
>something like the build Ceki put together is probably good enough to start with.  
>
>Has anyone brought this up on the log4j lists? 

No, but you are welcome to bring it up.


--
Ceki G�lc� - http://qos.ch

Reply via email to