Richard,

I am not a great big fan of lots of layers.  I would like to use the log API directly, 
that way I only have to know the *common* log api (which is already a layer) and I 
know how to use things.  It would be a little bit of sugar not to have to call the 
getMessage() inside every log call, but cut/paste and our unit test that enforces the 
requirement that we use getMessage() makes this not a problem in my book.  I would NOT 
be in favor of adding another logging layer.

I think getting a message APIs should live in the Message class, not the logging 
class.  Every class has a function...

I believe we have something (the JVM?) that enforces the rule that every key have an 
associate text in the properties file.  At least when I add a call without the text, 
things don't work for me, but what you propose (dumping the key with a warning) sounds 
reasonable.

--
Tom Jordahl
Macromedia Server Development



-----Original Message-----
From: Richard Sitze [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, September 18, 2002 12:41 PM
To: [EMAIL PROTECTED]
Subject: [axis] Messages & Logging : change direction?


I'd appreciate more than a silent consent from Tom & Glen :-)

Given the changes that are being discussed for Messages, now would be a 
GREAT time to:

- introduce a layer on top of logging that accepts message keys directly: 
I18NLog 
- let I18NLog perform translation
- expose getMessage() from I18NLog instead of Messages (this is the ONE 
piece I'm a bit uncomfortable with . . .) for other xlations (exceptions)
- if key doesn't map to text, then just dump key (issue debug/warning)?

Note that we already pass the Class name (i.e. package name) to 
LogFactory, and to make Messages aware of the class name we would be doing 
the same. . .

Thoughts?

<ras>

*******************************************
Richard A. Sitze
IBM WebSphere WebServices Development

Reply via email to