Additional debug loggers ?

[Brent Putman]

On 1/28/15 2:44 PM, Cantor, Scott wrote:
> I think it's ok within reason, but I don't want to have a general policy 
> of using these kinds of loggers in place of class loggers.

I was sort of already second-guessing the SAML_PRE_ENCRYPTION one.  My
main reason for doing that originally was that the Encrypter emits other
stuff on DEBUG.  But turns out it's really only a few things, so maybe
better to collapse it back in to the class logger.  The Encrypter is an
-api class not an -impl class, so don't think there's an issue with the
class itself.

Does anyone else have a strong opinion for or against?

>  If we really 
> thought that was necessary, then I would advocate revisiting the general 
> issue of what the categories should be in the classes.
>

Yes. I mentioned this is in the issue.  I imagine the vast majority of
our logging comes out of impl classes.  If we really think that is a
problem, we need to invent a whole mess of conceptual categories.

Personally, at least for DEBUG, I don't think that's necessary.  DEBUG
inherently means you are debugging the software, pretty low-level, and
for practical reasons I think that just means -impl classes.  So I
personally don't think having DEBUG logging config based on impl class
categories is a big problem.  For things that we want to specify as
consistent targets or something, that won't change if impl classes
change, then maybe conceptual categories make sense.  But then it's a
question of what/when/where, and for example whether to mix the 2
approaches in the same class, etc.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://shibboleth.net/pipermail/dev/attachments/20150128/345802f6/attachment.html