Hi Pavel,
On 16/02/17 13:37, Pavel Bucek wrote:
> Hi Sergey,
>
> please see inline.
>
>
> On 16/02/2017 12:04, Sergey Beryozkin wrote:
>> Hi Pavel,
>>
>> On 16/02/17 08:57, Pavel Bucek wrote:
>>> Hi Sergey,
>>>
>>> Flow.Subscription is going away. That part of the API is a sample of
>>> how things will look like when 2.1 is released - we will remove as
>>> much as possible from Flow.*. If there is something which cannot be
>>> removed (like Subscription), we'll consider the usecase (SSE in this
>>> case) and copy it into JAX-RS somehow (SseSubscription).
>>>
>> Sorry, so Flow.Subcription can not be removed ? Or if it is going
>> away and SseSubscription has been introduced, why did you decide to
>> keep it in m04 ?
>> When we discussed it with Andriy I thought the idea behind a new
>> SseSubscription was to encapsulate the JAX-RS specific 'Flow'
>> references from the rest of the API by inheriting from
>> Flow.Subscription so that when Java9 is there and JAX-RS Flow
>> eventually goes then the SSE API is not really affected. But I'm not
>> sure I follow now what is the idea behind keeping both
>> Flow.Subcription & SseSubscription in m04
>
> I don't want to remove Flow.* yet, because it will be used in future
> work, namely Non-blocking O/I.
>
> The original intention (as stated when Flow introduction email was
> sent to this list) was to keep it as long as possible but also with
> EDR submission approaching, we wanted to see how some part of the API
> will look when JAX-RS 2.1 is released. SseSubscription is one of the
> outcomes, #subscribe(...) method and overloads on SseEventSource is
> another outcome.
>
> I guess I could apply the same pattern to the Broadcaster and
> EventSink now and leave Flow unused (or remove it now); but what if we
> decide to use different pattern when finalizing NIO support? Then the
> transition will be even more difficult and would require more changes,
> since finalizing NIO would change SSE as well..
>
Lets just wait till 2.1 gets closer and revisit what stays and what not,
sounds good to me
>>> I know that it feels odd (and it is) to have this done on part of
>>> the API and not on another part, but it was part of the internal
>>> review process and my arguments were ignored..
>>>
>> I've no problems with the internal reviewers having their input but
>> I'd have problems with them, possibly not being JAX-RS 2.1 experts
>> affecting how the final 2.1 API will look like. I'm not keen to make
>> a big deal out of it, the proposed API in general is of high quality,
>> but well, I guess the final details should be finalized
>> in this group :-)
>
> well.. we do code reviews and we are not going to change that. I
> admitted that, but if I wouldn't (and now I'm thinking it was a
> mistake), you don't have a chance to get that info.
>
> The reviewer in this case was the original author of this part (and
> others) of JAX-RS API, Marek. That was also communicated out to the
> EG, see
> https://java.net/projects/jax-rs-spec/lists/jsr370-experts/archive/2017-02/message/43
>
> I always respond to any comment from the EG member, JCP member, random
> users_at_jax-rs-spec.java.net mail, random tweet, etc. Everyone who wants
> has the chance to influence the API. Of course, the feedback from the
> EG has higher priority and importance. Also, please understand that we
> need some feedback. It's hard to design things without a need to
> defend the API. The process usually brings up other possible
> approaches, which is what happened in the last case.
>
> I hope you are not suggesting to make the process open exclusively for
> EG members. If not, please let's drop this topic, it is not leading
> anywhere.
>
I definitely did not suggest it. You did not make a mistake, you said
earlier there was an internal review, it is not a secret and as I
implied the review was generally helpful.
However I'm against having a situation where here at the experts list we
come up with the idea that something needs to be changed and you will
reply, well, hard luck, the internal reviewers disagree. Hope you see
what I mean
Thanks, Sergey
> Best regards,
> Pavel
>
>>> Is this clearer now?
>> Getting there :-)
>>
>> Thanks, Sergey
>>
>>>
>>> Thanks and regards,
>>> Pavel
>>>
>>>
>>> On 15/02/2017 22:17, Sergey Beryozkin wrote:
>>>> Hi Pavel
>>>>
>>>> The question from my colleague Andriy (more to come)
>>>>
>>>> Can you clarify why both Flow.Subscription and SseSubscription are
>>>> both used interchangeably ?
>>>>
>>>> Thanks, Sergey
>>>>
>>>> On 15/02/17 15:47, Pavel Bucek wrote:
>>>>> Dear experts,
>>>>>
>>>>> if you have any feedback for SSE, please provide it by the end of
>>>>> this week, sooner the better.
>>>>>
>>>>> (it doesn't mean that feedback received afterwards won't be
>>>>> evaluated; it's mainly about keeping the ball rolling).
>>>>>
>>>>> Also, I'd like to announce that Early Draft Review will be started
>>>>> shortly afterwards. I believe we reached the point when we can say
>>>>> that we have done some progress and we could use feedback from
>>>>> wider group. EDR will hopefully help with that.
>>>>>
>>>>> Thanks and regards,
>>>>> Pavel
>>>>>
>>>>
>>>
>>
>