What about if we just use legacy extensions that would be loaded only on DC?
for legacy i mean, why not just have modules / jars from 7.2 in 8.0 distro?
that would make it easiest to support, and no extra work.
We should just put them in some special place in distro,
so it would be obvious that is legacy stuff only DC uses...


On Fri, Mar 1, 2013 at 11:44 PM, Brian Stansberry <brian.stansberry@redhat.com> wrote:
The extension registration logic would have to be altered to not barf
when multiple aliases all try to register the same extensions/ subsystems.

But it probably should still barf if a user tried to do that for some
other reason. So which is happening needs to be clarified.

A way to do that is to use something other than
org.jboss.as.controller.Extension for the ServiceLoader (i.e. first try
ServiceLoader for "LegacyExtension" and then if not there try for
org.jboss.as.controller.Extension.) That's hacky though unless there is
a real difference in the service API between Extension and what these
legacy extensions do. AFAICT though, there is no API difference;
difference is only in impl.

On 3/1/13 4:23 PM, David M. Lloyd wrote:
> Rewinding the discussion a bit :)
>
> If we just had one compat module (with N pure aliases), it could easily
> register all the subsystems for all the modules at that time (subsystem
> registration is pretty lightweight these days, or so it seems at a
> glance).  If extra subsystems are available as a result of an extension
> reg I don't see that as harmless.
>
> On 03/01/2013 02:48 PM, Brian Stansberry wrote:
>> I'm not sure how the ServiceLoader part would work there. At least not
>> with what I imagine when I think of an "alias." With some kind of stub
>> where each has a different
>> META-INF/services/org.jboss.as.controller.Extension file it could work.
>>
>> On 3/1/13 2:29 PM, David M. Lloyd wrote:
>>> Yeah, I was thinking they could just be aliases or stubs though.
>>>
>>> On 03/01/2013 02:22 PM, Brian Stansberry wrote:
>>>> In terms of code organization, perhaps. But the way the extension is
>>>> activated in the HCs and servers is via the module name. So if you want
>>>> a 7.2 server to be able to run CMP, there is going to have to be a
>>>> module named org.jboss.as.cmp.
>>>>
>>>> On 3/1/13 2:13 PM, David M. Lloyd wrote:
>>>>> I wonder - should we retain a skeletal version of each of these modules?
>>>>>        I was thinking maybe it would be better to maintain one big
>>>>> "removed-subsystems" or "compat-subsystems" module or something like
>>>>> that where we can neatly/consistently organize all the model stuff for
>>>>> these removals.
>>>>>
>>>>> On 03/01/2013 09:39 AM, Brian Stansberry wrote:
>>>>>> Thanks Thomas, for raising this and for the JIRA.
>>>>>>
>>>>>> I've outlined what I think is needed for the stub extensions as a
>>>>>> comment on https://issues.jboss.org/browse/AS7-6656 .
>>>>>>
>>>>>> Can I request that folks hold up on deleting these subsystems? I think
>>>>>> it will be easier to make these changes and then delete the unneeded
>>>>>> runtime stuff than it will be to semi-restore from history and then change.
>>>>>>
>>>>>> The ones that have already been deleted, it's no big deal.
>>>>>>
>>>>>> On 2/28/13 10:35 AM, Thomas Diesler wrote:
>>>>>>> Ok, stub extensions is the obvious alternative to breaking compatibility. I'll leave this as a future task and create a jira for it if that's ok with you.
>>>>>>>
>>>>>>> cheers
>>>>>>> --thomas
>>>>>>>
>>>>>>> On Feb 28, 2013, at 4:22 PM, David M. Lloyd <david.lloyd@redhat.com> wrote:
>>>>>>>
>>>>>>>> On 02/28/2013 05:57 AM, Thomas Diesler wrote:
>>>>>>>>> Folks,
>>>>>>>>>
>>>>>>>>> related to
>>>>>>>>>
>>>>>>>>> * [AS7-6612 <https://issues.jboss.org/browse/AS7-6612>] Remove JAXR support
>>>>>>>>>
>>>>>>>>> I'd like to know whether we need to preserve backward compatibility of
>>>>>>>>> the configuration and if so what should happen if there is a jaxr config
>>>>>>>>> item? Generally, can AS8 break backward compatibility with respect to
>>>>>>>>> the config?
>>>>>>>>
>>>>>>>> Brian points out that we don't have a specific requirement to maintain
>>>>>>>> compatibility with obsolete subsystems.  I think we could go ahead with
>>>>>>>> the removal (granted part of the reason I feel this way is that I've
>>>>>>>> already removed JSR-88...).
>>>>>>>>
>>>>>>>> Going forward though Kabir suggested that if we do want to, say, allow
>>>>>>>> 7.x instances to be managed from an 8.x DC, that we should create "stub"
>>>>>>>> extensions for the removed stuff that only carry and validate
>>>>>>>> configuration but aren't actually supported on 8.x servers.  This seems
>>>>>>>> like a valid possibility to me.
>>>>>>>> --
>>>>>>>> - DML
>>>>>>>> _______________________________________________
>>>>>>>> jboss-as7-dev mailing list
>>>>>>>> jboss-as7-dev@lists.jboss.org
>>>>>>>> https://lists.jboss.org/mailman/listinfo/jboss-as7-dev
>>>>>>>
>>>>>>> xxxxxxxxxxxxxxxxxxxxxxxxxxxx
>>>>>>> Thomas Diesler
>>>>>>> JBoss OSGi Lead
>>>>>>> JBoss, a division of Red Hat
>>>>>>> xxxxxxxxxxxxxxxxxxxxxxxxxxxx
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> jboss-as7-dev mailing list
>>>>>>> jboss-as7-dev@lists.jboss.org
>>>>>>> https://lists.jboss.org/mailman/listinfo/jboss-as7-dev
>>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>
>>>
>>
>>
>
>


--
Brian Stansberry
Principal Software Engineer
JBoss by Red Hat
_______________________________________________
jboss-as7-dev mailing list
jboss-as7-dev@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/jboss-as7-dev