> On 20 Apr 2015, at 02:24, Jacques Le Roux <jacques.le.r...@les7arts.com> 
> wrote:
> 
> Le 20/04/2015 09:47, Adrian Crum a écrit :
>> Generally speaking, I am in favor of using another framework. I have two 
>> reservations about Moqui:
>> 
>> 1. It is controlled by a single person - so responsiveness to issues are 
>> dependent on that person's availability.
> 
> This is indeed a regression from the current community sharing. On the other 
> hand a such change would not be done in one day, so we would have a long 
> period to experiment in parallel before possibly switching to Moqui.
> I also guess in such cases David could open the Moqui to people he trusts. I 
> though wonder how this would be linked together. Nothing blocking but to be 
> seriously thought about, not only technically but legally. I know it's 
> David's will to share and he proved it already with OFBiz but the licensing 
> aspect is not clear to me http://www.moqui.org/#model.

With Moqui it's no secret that I chose the "code over community" route which is 
certainly different from the ASF emphasis on community (and my emphasis on 
community and collaboration from the beginning in OFBiz). 

Moqui Framework is now pretty mature and I have sent out a couple of 
solicitations (in the LinkedIn group) for more contributors, but I'm in no 
hurry to bring on other committers/moderators... better to wait until clearly 
interested and capable people come along. The infrastructure is in place for 
this on GitHub, ie the main repo is under the "moqui" group and not my personal 
account any more so it is easy to add others to that group with permission for 
particular repositories (ie moqui vs mantle, HiveMind, etc).

Part of the reason I'm less worried about this is the different model for 
source management that Git makes possible and GitHub makes easy. In other 
words, the magic of distributed source management. Distributed source 
management and the moderator model make it possible for "forks" to exist that 
have variations on the main code base that are available for all and pulled 
into the main code base by moderators. You don't have to be a moderator to 
contribute, or even share your code with the world.

The whole distributed source and moderator model is very different from the 
community model in OFBiz, and overall better for some things and worse for 
others. Many big projects use this model, including the Linux kernel which is 
probably the biggest and the use of this model for the Linux kernel is exactly 
where git came from.

My guess is that if OFBiz started using Moqui Framework one or more current 
OFBiz committers would become moderators with commit access to the main Moqui 
Framework repository.

>> 2. It repeats a lot of mistakes that have been made in OFBiz, so those 
>> things will need to be fixed again in Moqui after we bring it on board.
> 
> Indeed, a lot of fixes have been done recently in OFBiz which is battle 
> tested for years. Moqui though certainly well done, is still young and we 
> would need to compare them, point by point.

I'd love to hear more detail on this. I know one criticism in the past from 
Adrian is Moqui's approach to object type conversions, especially in the Entity 
Facade (the part of Moqui similar to the OFBiz Entity Engine). In some parts of 
Moqui, like for service parameters, it uses the Groovy type conversion which is 
pretty good but isn't a pluggable type conversion framework as I'm sure Adrian 
would like to see... but that honestly I still have yet to find a need for.

On the entity level I prefer the current approach, mostly for performance 
reasons. In recent changes to the framework for performance improvements (which 
were significant... Moqui now getting around 15,000 entity ops per second as 
opposed to 3,000 before the latest changes, and close to 300 in the original 
version of Moqui though that was on an older laptop/etc and before ANY 
optimizations where done). This does limit the types available for entity 
fields, but is that really an issue? You're going to a database through JDBC, 
and really just the least common denominator of types available in databases to 
cleanly run on various ones, so there aren't many options in the first place! 
We can take advantage of that for simpler code and better performance.

>> Neither one is a show-stopper for me.
> 
> Same here, just cautious.

Caution is good in this case... to be honest I'm not sure using Moqui Framework 
in OFBiz is a good idea (as I've mentioned before), will write more on that in 
another reply.

-David


Reply via email to