On Nov 12, 2012, at 9:21 AM, Olivier Lamy <[email protected]> wrote:

> 
> Currently I'm testing integrating jansi to have colorized output, that
> works fine and that's fun :-)
> Again I don't see why we couldn't add a bit or a possibility of fun
> within our distribution (or at least make that easily possible)

Sure, pretty easy with Logback: 
http://logback.qos.ch/manual/layouts.html#coloring

If that is really the only feature then there is another provider that just 
does the coloring: https://github.com/jdillon/gossip

But again really, I believe the decision is to determine whether all this 
really necessary. I think the order of operations that folks think is 
reasonable is:

1) Get the release out with SLF4J
2) Determine whether we need logging framework features
3) Pick one

That's what I've gathered from the discussion so I think if we plan that route 
then we can do 3.1.0, and then have the discussions and if we get to 3) maybe 
integrate that into 3.1.1 or 3.1.2.

Sound reasonable. 

>> 
>> Then we should ship this in its current form, discuss whether we need 
>> advanced logging, and then look at the implementations. I have one using 
>> Logback and I think that solution is more mature, and has a community and 
>> used heavily, even by many other Apache projects. I looked at Log4j2 and 
>> there are 2 people essentially (and one fellow with 2 commits) and I'm not 
>> sure how the project started but I don't think it even passes Apache 
>> Incubator standards. At first blush I don't see log4j2 as a good choice. 
>> Hence, I agree, a discussion is required. But I think we might arrive at the 
>> conclusion a logging framework is not necessary.
>> 

> IMHO the number of contributors is not a good argument.

Let's call it the ecosystem factor then. I consider users and contributors the 
most important there. As a library is nothing without users because a library 
with users generally has developers even if the library has reached maturity. A 
library with only developers and not many users isn't very valuable overall as 
there is really no ecosystem.

> How many people really contributed to sisu

There is only one alternative in the world for Sisu and that's Plexus. But at 
the core of Sisu is Guice which has a vast ecosystem. I don't think anyone will 
disagree it was a good choice to get off Plexus and use Sisu with the emulation 
layer it provides. Guice has had 5-10 developers at any given point in time and 
the technology is evolving in the form of Dagger by Square. So Sisu is unique 
in what it does, but Guice is not. We picked a solid base with a good ecosystem 
to build on. Maven 3.x is actually a testament to how flexible Guice is and how 
responsible they were to absorbing our many changes to make it work. None of 
this would have worked without a lot of work and help from many people. Guice 
is mature, has lots of users, actively developed and evolving.

In this case we are talking about the exact opposite. We have the most mature 
SLF4J provider with a history, users, lots of forks and pull requests and is 
pretty much becoming the de facto standard. On the other side you have a 
logging framework that no one really uses in comparison. There is a completely 
mature alternative.

> or aether ?

Again, what's the alternative to choose from? Both Aether and Sisu required a 
lot of specialized work for Maven. No one else really has time or energy to do 
these things. It's not commodity infrastructure like DI or logging.

> And those
> libraries are more core part of maven than a logging framework.
> What you call integrators can easily change it (as core will only use
> slf4j api and no framework specific except maybe some sys props as
> it's already the case with your changes using slf4j-simple) but AFAIK
> that's not possible for those libraries !
> I just have a look at the impact graphs in github and frankly not a
> lot of people contributed to those libraries.
> 
> At least with log4j2 we will have a framework maintained by the Apache
> community so I think number of contributors will grow..
> 

That's speculation and empirically from the stats that doesn't look to be the 
case:

http://svnsearch.org/svnsearch/repos/ASF/search?path=%2Flogging

Ceki is still the largest contributor in that community and he's not there 
anymore.

I'm personally not in favor of giving an unproven library like log4j2 
undeserved legitimacy by being used in Maven when Logback exists. No one is 
working on log4j really anymore so how can you say that work will continue with 
log4j2? And the only guy really still working on log4j is not really working on 
log4j2?

http://svnsearch.org/svnsearch/repos/ASF/search?path=%2Flogging%2Flog4j%2Ftrunk

I just don't see using log4j2 making much sense.

At any rate, can we agree on the plan above and table the discussion of whether 
we need a framework and the possibly selection of one until 3.1.1/2? That gives 
us time to discuss and create example implementations if the outcome of the 
discussion is to move forward with framework selection.

Thanks,

Jason

----------------------------------------------------------
Jason van Zyl
Founder & CTO, Sonatype
Founder,  Apache Maven
http://twitter.com/jvanzyl
---------------------------------------------------------

You are never dedicated to something you have complete confidence in.
No one is fanatically shouting that the sun is going to rise tomorrow.
They know it is going to rise tomorrow. When people are fanatically
dedicated to political or religious faiths or any other kind of 
dogmas or goals, it's always because these dogmas or
goals are in doubt.

  -- Robert Pirzig, Zen and the Art of Motorcycle Maintenance





Reply via email to