Re: svn commit: r734201 - in /webservices/axis2/trunk/java/modules: addressing/pom.xml saaj/pom.xml

[Glen Daniels <gl...@thoughtcraft.com>]

Hi Asankha, all:

Asankha C. Perera wrote:
> Hi Jarek
>> I agree with changing the transport dependencies to test scope but in
>> general I'm a little concerned about the circular dependency between
>> axis2 and axis2-transport modules. Because of that circular dependency
>> I think we have to release axis2 and axis2-transport at the same time
>> (at least the base, local, tcp and http transport modules) and I
>> thought the idea behind splitting transport code from Axis2 was to
>> release the transport modules independently from Axis2.
>>   
> Yes, for some reason some of the existing pom.xml's creates problems for
> projects down the line, with these issues. Its important to set the
> dependencies for runtime and testing etc correct. Also to fix the
> dependencies to avoid any circular dependencies. I firmly believe that
> transports need to be released separately, and its just a matter of time..

+1.  We've done the first step of factoring the transports, but not the
whole enchilada.  The big philosophical change we now need to make is
that Axis2 shouldn't, aside from test code, be relying on any particular
set of transports (except *possibly* the local one).  It would also be
really great if we could optionally run most of the Axis2 tests which
currently use HTTP over the local transport - I would think that might
speed up the build significantly.

For Axis2 1.5, we DO need to get a real (non-SNAPSHOT) release of the
transports out ASAP.  Will bring that up on commons-dev.

--Glen

Re: svn commit: r734201 - in /webservices/axis2/trunk/java/modules: addressing/pom.xml saaj/pom.xml

[Jarek Gawor <jg...@gmail.com>]

On Wed, Jan 14, 2009 at 12:37 PM, Glen Daniels <gl...@thoughtcraft.com> wrote:
> Hi Asankha, all:
>
> Asankha C. Perera wrote:
>> Hi Jarek
>>> I agree with changing the transport dependencies to test scope but in
>>> general I'm a little concerned about the circular dependency between
>>> axis2 and axis2-transport modules. Because of that circular dependency
>>> I think we have to release axis2 and axis2-transport at the same time
>>> (at least the base, local, tcp and http transport modules) and I
>>> thought the idea behind splitting transport code from Axis2 was to
>>> release the transport modules independently from Axis2.
>>>
>> Yes, for some reason some of the existing pom.xml's creates problems for
>> projects down the line, with these issues. Its important to set the
>> dependencies for runtime and testing etc correct. Also to fix the
>> dependencies to avoid any circular dependencies. I firmly believe that
>> transports need to be released separately, and its just a matter of time..
>
> +1.  We've done the first step of factoring the transports, but not the
> whole enchilada.  The big philosophical change we now need to make is
> that Axis2 shouldn't, aside from test code, be relying on any particular
> set of transports (except *possibly* the local one).  It would also be
> really great if we could optionally run most of the Axis2 tests which
> currently use HTTP over the local transport - I would think that might
> speed up the build significantly.

A couple of comments. First, I think it is important to run tests
(some at least) over HTTP to tests stuff like session support,
cookies, response codes, etc. All of which are important from spec and
spec compliance point of view. Second, the transport dependencies in
Axis2 are not just needed in tests. We include the transport modules
in the Axis2 web application and the binary distribution. And after
somebody installs Axis2 I don't think we are going to expect them to
also download and install the basic transport jar files just to make a
simple Axis2 client working. So, I would argue that the http transport
module (and whatever it requires and maybe local and tcp) should move
back into Axis2 because they are the minimal set of transport modules
that make Axis2 work. That would also eliminate the circular
dependency problem that we have today and let other Axis2 transport
modules to be released independently from Axis2.

Jarek