Hi Pavel,

 

Thank you for the clarification, and my apologies for bringing this up so late - a colleague and I ran into some confusion on this point, so I thought I would bring if forward in hopes that other users might avoid a similar confusion.

 

I have opened a pull request [1] with my suggested changes to the javadoc.  I think this may help users better understand the differences in the Java SE vs EE environments.

 

Thanks again - it has been a pleasure working with you in this expert group!

 

Andy

 

[1] https://github.com/jax-rs/api/pull/560

J. Andrew McCright
IBM WebSphere Development
+1 507 253 7448
TL 553-7448
andymc@...

 

 

----- Original message -----
From: "Pavel Bucek" <pavel.bucek@...>
Sent by: jaxrs-spec@javaee.groups.io
To: jaxrs-spec@javaee.groups.io
Cc:
Subject: Re: [jaxrs] Proposed Final Draft
Date: Fri, Jul 7, 2017 12:25 AM
 

Hi Andy,

it is very late for changing anything..

We are not saying that on Java EE environment ClientBuilder#executorService and #scheduledExecutorService have no effect - but it might sound that way.

The "default" for Java SE is NOT ForkJoinPool#commonPool, since it is not meant for I/O operations. On Java SE, the default is intentionally left as implementation specific.

And I'm not sure what you mean by "what about Future Objects returned from ..." - if there is any need for starting a thread, it should be done on the executor service provided in the client builder. Note that the javadoc for ClientBuilder#executorService mentions:

* Provided executor service will be used for executing asynchronous tasks.

And Invocation#buildGet() (and others):

* Submit the request for an asynchronous invocation and receive a future
* response back.

Which seem to imply what I wrote..

It might be still possible to tweak the javadoc, I'll wait for your reaction - whether this makes sense or not and then will inquire about possible javadoc adjustment.

Thanks and regards,
Pavel

On 07/07/2017 00:35, Andy McCright wrote:

Hi Pavel,

 

If it is not too late, I'm wondering if we could clarify the uses of ClientBuilder.executorService(...) and scheduledExecutorService(...).

 

Are we saying that in a Java EE environment, these methods have no effect?  If so, can we change this text:

When running in a Java EE container, implementations are required to use the container-managed executor service.

 

to:

When running in a Java EE container, this method will have no effect.  The container will always use the container-managed executor service.

 

and a similar change for the executorService method?

 

 

If instead we are saying that in a Java EE container, the default ES and SES are different than in Java SE, we should clarify it like this:

When running in a Java EE container, the default scheduled executor service will be the container-managed service.  In Java SE, this will be the ForkJoinPool service.

 

 

I think we should also clarify what methods will use these services.  The javadoc currently provides some guidance with:

     * @see Invocation.Builder#async()
     * @see Invocation.Builder#rx()
     * @see RxInvokerProvider#getRxInvoker(SyncInvoker, ExecutorService)

and

     * @see SseEventSource.Builder#reconnectingEvery(long, TimeUnit)

 

but what about Future objects returned from methods like buildGet().submit(...)?

 

Thanks,

 

Andy

 

J. Andrew McCright
IBM WebSphere Development
+1 507 253 7448
TL 553-7448
andymc@...

 

 

----- Original message -----
From: "Pavel Bucek" <pavel.bucek@...>
Sent by: jaxrs-spec@javaee.groups.io
To: jaxrs-spec@javaee.groups.io
Cc:
Subject: [jaxrs] Proposed Final Draft
Date: Thu, Jun 22, 2017 6:18 AM
 

Dear experts,

we'd like to announce that Proposed Final Draft was published on JCP JSR
370 page [1].

I'd like to encourage you all to review provided artifacts and let us
know if you have any feedback.

Thanks and regards,
Pavel & Santiago

[1] https://jcp.org/en/jsr/detail?id=370