users@grizzly.java.net

Re: Grizzly ROADMAP: 1.9.x and 2.0 -> Time to speak!

From: Jeanfrancois Arcand <Jeanfrancois.Arcand_at_Sun.COM>
Date: Fri, 16 Oct 2009 16:56:45 -0400

Salut,

Matthew Swift wrote:
> Hi,

Apology for the delay
>
> As new users of Grizzly and having had no experience of previous
> releases, I thought I'd share our initial impressions and
> recommendations for 2.0. I don't know how much they make sense because
> our experience is pretty restricted to using Grizzly as a pure IO
> framework, and we don't have much interest in the Web related related
> functionality (e.g. HTTP / Comet) at the moment.
>
> So here goes: our first impression is that it's very easy to get lost in
> the APIs and not know where to start (this may be a direct result of the
> documentation being currently aligned on 1.9.x releases?). There seem to
> be some key classes buried down in sub-packages and also overlapping
> functionality (e.g. codecs vs filters).

Honestly I haven't yet started working on 2.0, letting Alexey driving
it. I will for sure pay attention to the above comment when I start.

For sure I want a back to back white paper included in that release. We
never did it for grizzly 1.9.x and it was a big mistake. The growing
popularity of Netty is an indication that great API + Full documentation
is mandatory.

Another thing that bugs me is
> that everything is public which adds to the confusion. Therefore I think
> that some time should be spent on the following:
>
> * Update/finish Javadoc and update documentation on website.

Agree

>
> * Make an effort to hide as much implementation detail as possible:
> only expose classes / methods / fields which are intended for
> public consumption. Make classes/methods final where possible and
> document things that are intended for extension. All basic Joshua
> Bloch stuff really. :-)

:-)


>
> * All aspects of performance and scalability are absolutely critical
> for us, but we are not yet at a point where we can provide useful
> feedback (our Grizzly based prototype is not complete yet). Our
> initial impressions are good though :-)

For sure we will do the usual stress test we did for 1.0 and 1.5 when we
started. We have an entire performance team that track us closely.


>
>
> Regarding Github vs Kenai and (and SVN vs other VCS). Being Java.net
> based ourselves we understand the pain you are experiencing! :-( Before
> migrating away I would consider the following:
>
> * Where do you plan to put the website content, mailing lists, WIKI,
> and bug tracking? I don't think Github provides support for any of
> these from a first glance, although I could be completely wrong.
>
> * User migration / authentication: I wouldn't be surprised if
> Project Kenai provides a migration path for users from java.net.
> It's probably worth checking.

Ha I didn't know that one. Good reason.


>
> * SVN vs Git vs Hg, etc: personally I would stick with SVN as it's
> fine for most needs, is easy to use, and has very good IDE support
> in Netbeans, Eclipse, and IntelliJ. Our experience of using SVN
> and Hg internally is that there is no noticeable performance
> difference (probably the same applies for Git but I'm just
> guessing) - the reason SVN is slow for us is more to do with
> Java.net than anything else. So I suggest changing provider first
> before changing VCS.

Agree, except SVN merge are my nightmare :-) That's probably the only
reason (valid for me).

Thanks!!!

-- Jeanfrancois

>
> Matt
>
> ---
>
> OpenDS Project: http://www.opends.org
>
>
>
>
>
>
> On 10/08/2009 05:48 PM, Jeanfrancois Arcand wrote:
>> Salut,
>>
>> Now that we are *almost* saved from GlassFish work (almost :-)), we
>> will eventually have time to come back here and do what we like,
>> which is to make the community happy and add new features. I would
>> like to start a new thread discussions about where peoples want to
>> focus on.
>>
>> Hubert has done great work on the Deployer and Sebastien on annotation
>> support. What's next? I would propose we work on two things:
>>
>> (1) Grizzly 2.0 FCS, with a back to back white paper as a
>> pre-requisite for release. I sign up for the white paper :-)
>>
>> (2) Forward port all the http/http-servlet-XX/Comet/OSGi works to
>> Grizzly 2.0
>>
>> ** Anybody interested to contribute I will give committer privileges
>> for free :-) **
>>
>>
>> Now I would like to vote on the following topic:
>>
>> 1. Should the Servlet Container works be only on top of 2.0? Or should
>> we make sure both version are supported?
>>
>> 2. Should 1.9.19 be *our last* 1.9.x release with new features, and
>> then we move to sustaining mode?
>>
>> 3. Should we refactor *entirely* the http module API and make sure it
>> is aligned with Grizzly 2.0 new design? I'm tempted to refactor
>> everything but we will break a lot of applications
>>
>> 4. Gitub or Kenai? For sure we MUST move out of java.net
>>
>> 5. Should we drop Cometd support? Cometd.org can be run using Grizzly
>> Comet + Atmosphere
>>
>> If you can reply publicly, please reply to jfarcand_at_apache.org with
>> your request/recommendation.
>>
>> Please share your vision :-)
>>
>> A+
>>
>> -- Jeanfrancois
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: users-unsubscribe_at_grizzly.dev.java.net
>> For additional commands, e-mail: users-help_at_grizzly.dev.java.net
>>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe_at_grizzly.dev.java.net
> For additional commands, e-mail: dev-help_at_grizzly.dev.java.net
>