On 09/07/14 20:35, Brian Stansberry wrote:
On 7/9/14, 1:59 PM, Stuart Douglas wrote:
> The problem with that is that if you take it to its fullest extent you
> end up with one subsystem per repo, which is not something we want.
>
> I am not sure where the best place for it is, even if it stays in core
> it should be possible for the tooling to exclude it, same with logging.
>
> Otherwise I think the place for it to live would be the web distro, as I
> think that people will definitely want to be able to use JMX to manage
> that.
>
So the web build is becoming the spot where foundational stuff like this
and Elytron come in?
My view on Elytron is that the core is going to have to have a
dependency on the Elytron module, the actual APIs and SPIs of Elytron
are going to have to be tightly integrated in the core.
But the subsystem itself and any related additional providers e.g.
PicketLink / KeyCloak are things I would see added in the functional
distributions so yeah based on what Stuart said web probably becomes the
entry point for this.
The core is uber-minimal for the folks who really
want that, and then web has these things that lots and lots of folks
will want.
In the odd case where folks want this foundational stuff but not
undertow etc, they can just depend on the web build and exclude
undertow. Real corner case. And folks who don't want the foundational
stuff exclude it.
I can see that working out pretty well.
Does logging belong in web then then? Still seems like something that
even the uber-minimalists would want. I ask because it bugs me that we
have two meanings now for "core" -- the old "core" notion that was
the
true core with zero subsystems, and now this new wildfly-core dist,
which has subsystems.
>
>
>
> Fernando Nasser wrote:
>> Shouldn't it be like an onion?
>>
>> A more bare core. A Core with monitoring capabilities. And so on.
>>
>>
>> On 2014-07-09, 2:27 PM, Brian Stansberry wrote:
>>> My earlier reply on this thread was in the branch focused on the
>>> technical detail of how the dependency arises. Which should be fixed;
>>> sounds like that's happening.
>>>
>>> On the broader question of where stuff belongs, JMX is an interesting
>>> case. Lots of users will want JMX. A number of subsystems have optional
>>> dependencies on it, including some like jgroups and infinispan that may
>>> end up being part of their own dist, separate from the full dist.
>>>
>>> So what's the plan for catering to these needs? Ship it in the full dist
>>> and those who want it depend on the full dist but configure the
>>> provisioning tool to only grab the JMX modules? Make it a micro-dist?
>>> Maybe a mini-dist packaged with some others? JMX, system-jmx, pojo as
>>> the "deploy your own services" dist? (I don't like that last
one, just
>>> brainstorming...)
>>>
>>> On 7/9/14, 5:47 AM, Darran Lofthouse wrote:
>>>> Working with the split repo just questioning if JMX is really needed in
>>>> core?
>>>>
>>>> Whilst most distributions would include it I am not convinced it is a
>>>> subsystem all must have.
>>>>
>>> Manageable logging isn't either. Lots of people used logging.properties
>>> for a long time.
>>>
>>> Note I'm not advocating removing logging from core; I'm just saying
that
>>> no subsystem has to be in core if that's the criteria.
>>>
>>>> Regards,
>>>> Darran Lofthouse.
>>>> _______________________________________________
>>>> wildfly-dev mailing list
>>>> wildfly-dev(a)lists.jboss.org
>>>>
https://lists.jboss.org/mailman/listinfo/wildfly-dev
>>>>
>>
>> _______________________________________________
>> wildfly-dev mailing list
>> wildfly-dev(a)lists.jboss.org
>>
https://lists.jboss.org/mailman/listinfo/wildfly-dev
> _______________________________________________
> wildfly-dev mailing list
> wildfly-dev(a)lists.jboss.org
>
https://lists.jboss.org/mailman/listinfo/wildfly-dev
>