Hi,
I have thought some more about this and the following may be what is
required:
1) An AtmosphereJerseyServletContainer that extends or adapts
ServletContainer.
That servlet can be referred to in the web.xml.
The configure method of ServletContainer can be overridden to
modify the ResourceConfig.
2) Modify ResourceConfig so that filters can be added programatically.
Nice to have:
3) Declare Guice/Spring support using a context listener.
Because a specific servlet is declared we need another
mechanism to declare support rather than the current
mechanism of using Guice/Spring specific servlets. I think this
approach is useful anyway.
4) When underlying Servlet 3.0 support is available we could utilize
an @Atmosphere annotation on the
class that extends Application. This would be a meta-annotated
with an annotation that declares the
servlet in 2) to be used.
5) A provider mechanism to initiate ResourceConfig.
Paul.
On Sep 24, 2009, at 5:28 PM, Paul Sandoz wrote:
> On Sep 24, 2009, at 4:58 PM, Jeanfrancois Arcand wrote:
>>>>
>>>> Any solutions will works. Right now Atmosphere uses the
>>>> ServletContainer but that may change since we have discussed the
>>>> possibility of running Atmosphere on top of Jersey instead of
>>>> Jersey on top of Atmosphere.
>>>>
>>> Ah! topsy turvy :-) why?
>>
>> You didn't want to be able to support some annotation here instead
>> of having to define Atmosphere in web.xml? I was under the
>> impression that we discussed something like @Atmosphere support,
>> but here :-)
>>
>
> OK, i got it now. It was not clear to me that required Atmosphere on
> top of Jersey, but i see what you mean.
>
> IIRC the initial thought was can the number of XML configuration
> files be reduced to just the web.xml for web deployments.
>
> The annotation occurred to me if there was a class that extends
> javax.ws.rs.core.Application that it could be annotated with
> something like @Atmosphere. Thus, if using Servlet 3.0 one does not
> require a web.xml.
>
>
>
>>>>
>>>>> If so i think Victor's solution of a ResourceConfigTransformer
>>>>> (as sent on the atmosphere list, see attached) seems like a good
>>>>> solution.
>>>>
>>>> +1
>>>>
>>> Could you log the issue if not already, i don't recall if you did,
>>> but i am in the process of sensory information overload at the
>>> moment.
>>
>> Done.
>>
>
> Thanks,
> Paul.