Hi Chinthaka:
Eran Chinthaka wrote:
Any thoughts on why leave behind XmlSchema, Neethi, Axiom?
Because they are used by several other projects that have nothing to do
with Axis 2?
This is not a good argument, at least for me. We are using lots of
components, like mail implementation from Geronimo project. But are they on
Geronimo commons? Aren't we accessing them, even when they are within
Geronimo? Is it a must to have them separate?
It is not a must, no, which is why we're having this conversation. :)
However, the goal of this whole discussion is to reorganize things to
enable a) smaller, more focused PMCs, and b) better sharing of
components. If we just s/WS/Axis2/ and leave everything in one project,
then it's not really a worthwhile refactor, IMO. I also don't think we
should pull every individual component out into its own project -
modularity is an art not a science.
So whats the problem of moving Axiom, Neethi also in to Axis2 project. As
one said, a change in Axiom might most probably affect Axis2 as well.
So we should move StAX and Activation and Servlet into Axis2 also, then?
I don't think this logic holds.
There are two more reasons I have *for* moving Axiom also in to Axis2
project *if* Axis2 becomes a TLP.
1. Axiom has a SOAP implementation which was done because of Axis2. If one
needs to leave Axiom inside WS, during this movement, then some one please
move the SOAP part out of Axiom. If the argument is Axiom is a pure XML
object model so that others can use it, then please move out the SOAP part
from Axiom.
Please note that I'm not saying anything bad about Axiom having a SOAP
implementation. We did it because we targeted Axiom for Axis2.
Axiom's SOAP implementation is in no way tied to Axis2, and could be
used by any SOAP processor... no? I'm actually fine with the idea of
refactoring the SOAP part out into a separate jar, but I still think it
should live in WS/Axiom, not in Axis2.
2. In another thread, it was suggested to make POLOKA a sub-project of
Axis2, because
i) it *will* have Axis2 compatible eventing implementation
ii) and (may be) it is Web service based implementation. (please correct me
if I'm wrong)
If the argument for leaving some projects within WS project, and taking with
Axis2 is about Axis2 dependencies, then why should we move Poloka (when it
is graduated) in to Axis2. Its just another project, which has a *plugin*
(as I think) to work with Axis2.
if this is the case, then I will argue that Synapse also should come under
Axis2. Abdera should come in to Axiom.
I am seriously missing something somewhere.
No argument from me there. :) Those last couple of paragraphs were crazy!
Seriously though, the decision of where Poloka graduates will happen at
the appropriate time, and should depend on what the project looks like
at that point. If it's purely an Axis2 module, then IMO it should
absolutely go under Axis2. If it's a reusable framework for event
brokering that gets used by several projects (ServiceMix, etc), then it
should either go into WS (to be shared) or its own TLP if it's big enough.
The basic idea here is to avoid giant umbrella PMCs, focus communities
and effort, and make it easier to share work amongst multiple projects.
So your take on things is basically "it's not worth the bother",
Chinthaka? I can respect that, although I don't agree with it.
Thanks,
--Glen
