Re: [logging] JCL in SLF4J flavour - a demo for discussion

[Simon Kitching <sk...@apache.org>]

On Mon, 2007-03-26 at 15:51 +0200, Oleg Kalnichevski wrote:
> On Fri, 2007-03-23 at 19:56 +0100, Boris Unckel wrote:
> > Hello,
> > 
> > I have seen the recent discussions on JCL 2.0.0 and a version without 
> > autodiscovery.
> > Someone stated to stop any further development (with good reasons 
> > behind) but I am
> > thinking different.
> > 
> > Please have a look at the (working) code:
> > https://issues.apache.org/jira/browse/LOGGING-112
> > 
> > It has a static binding, some improvements in the logfactories 
> > (recognizes native implementations).
> > API and implementations are cleanly separated, the 1.1.x diagnostic 
> > function is still used.
> > 
> > What are your thoughts?
> > Is this a possible direction?
> > 
> 
> Hi Boris,
> 
> I think this patch represents the way I personally would like to see JCL
> evolve in the future. If JCL 2.0 were being developed along these lines,
> I (personally) would be very inclined to continue using JCL for the next
> major version of HttpClient.

Would you both mind explaining what benefits you see in a new JCL
implementation that cannot be obtained via java.util.logging?

I'm no fan of the j.u.l design, but it seems to me less effort to use it
than to fight it. What have I not understood?

And what would a new JCL do for anyone that they could not do by just
using SLF4J via its JCL API? The SLF4J team have already done a lot of
hard work; what benefit would there be to duplicating that?

Regards,

Simon


---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org

Re: [logging] JCL in SLF4J flavour - a demo for discussion

[Boris Unckel <bo...@gmx.net>]

Hello Simon,

Simon Kitching wrote:
> Would you both mind explaining what benefits you see in a new JCL
> implementation that cannot be obtained via java.util.logging?
>   
this is possible already today with x4juli, it does have a JCL native 
implementation.

> I'm no fan of the j.u.l design, but it seems to me less effort to use it
> than to fight it. What have I not understood?
>   
JCL apps/libraries are widely spread. A change in the interface or the 
public factory methods is
very unlikely. The difference between JCL 1.0.x, 1.1.x and JCL in SLF4J 
style is just the way
of packaging (which classes are in the different jars) and the binding 
(without autodiscovery,
currently (to be discussed) without classloader support). Another 
difference is the check and use for
native implementations.

One last point: The extension of JUL always requires classes on the 
system classpath. This is especially
annoying if you have application specific dynamic filters in a container 
because you cannot deploy everything
together, you must change the container config.
> And what would a new JCL do for anyone that they could not do by just
> using SLF4J via its JCL API? The SLF4J team have already done a lot of
> hard work; what benefit would there be to duplicating that?
The benefit is to have the same way of control as SLF4J. Maybe the need 
for bridging is heavily
reduced. I personally do not like the JCL -> SLF4J -> 
CurrentUnderlyingLibrary or the
SLF4J -> JCL -> CurrentUnderlyingLibrary way. It seems better to me to 
have both wrappers
directly sending to the underlying library.

This code has been presented to be discussed, to get away from 
theoretical thoughts without
any prove of concept or similiar - thanks for participating the discussion.
What is your prefered way of underlying library discovery?

Regards
Boris


---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org

Re: [logging] JCL in SLF4J flavour - a demo for discussion

[Oleg Kalnichevski <ol...@apache.org>]

On Tue, 2007-04-03 at 22:51 +1200, Simon Kitching wrote:
> On Mon, 2007-03-26 at 15:51 +0200, Oleg Kalnichevski wrote:
> > On Fri, 2007-03-23 at 19:56 +0100, Boris Unckel wrote:
> > > Hello,
> > > 
> > > I have seen the recent discussions on JCL 2.0.0 and a version without 
> > > autodiscovery.
> > > Someone stated to stop any further development (with good reasons 
> > > behind) but I am
> > > thinking different.
> > > 
> > > Please have a look at the (working) code:
> > > https://issues.apache.org/jira/browse/LOGGING-112
> > > 
> > > It has a static binding, some improvements in the logfactories 
> > > (recognizes native implementations).
> > > API and implementations are cleanly separated, the 1.1.x diagnostic 
> > > function is still used.
> > > 
> > > What are your thoughts?
> > > Is this a possible direction?
> > > 
> > 
> > Hi Boris,
> > 
> > I think this patch represents the way I personally would like to see JCL
> > evolve in the future. If JCL 2.0 were being developed along these lines,
> > I (personally) would be very inclined to continue using JCL for the next
> > major version of HttpClient.
> 
> Would you both mind explaining what benefits you see in a new JCL
> implementation that cannot be obtained via java.util.logging?
> 
> I'm no fan of the j.u.l design, but it seems to me less effort to use it
> than to fight it. What have I not understood?
> 
> And what would a new JCL do for anyone that they could not do by just
> using SLF4J via its JCL API? The SLF4J team have already done a lot of
> hard work; what benefit would there be to duplicating that?
> 
> Regards,
> 
> Simon
> 

Let me take my best shot at trying to explain it.

(1) HttpClient is meant to be a low level transport library and as such
it is supposed to be as non-intrusive toward its upstream dependencies
as possible. To me _personally_ this also means not imposing a
particular choice of a logging toolkit. The reliance on JUL for logging
in HttpClient 4.0 would seem to prevent the use of some popular
toolkits. I am _personally_ not aware of a good JUL to Log4J adapter

(2) I _personally_ am fine with just dumping JCL in favor or SLF4J, but
other HttpComponents committers are currently not.

(3) The only serious gripe about JCL I am aware of is its auto-discovery
mechanism. There _appears_ to be a consensus, please correct me if I am
wrong, that auto-discovery was indeed a mistake.    

(4) I _personally_ still have a rather naive belief that Jakarta stands
for something and we should "eat our own dog food" whenever possible. If
JCL 2.0 were to address the auto-discovery issue, that would remove my
only reservation about using JCL in HttpClient 4.0.

That's just my take on the issue for what it is worth. 

Oleg

> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: commons-dev-help@jakarta.apache.org
> 
> 


---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org