[jsr294-modularity-eg] Simple Module System Delema

[Stephen J. McConnell]

On Thu, 2009-09-17 at 18:02 +0100, Neil Bartlett wrote:
> Stephen,
> 
> On Thu, Sep 17, 2009 at 4:39 PM, Stephen J. McConnell
> <mcconnell at dpml.net> wrote:
> > Also, while I notice frequent references to SMS, I am very much aware of
> > a complete absence of any reference to any formal documentation outside
> > of the Jigsaw project page links (keeping in mind that the 294
> > referenced content is so old it's somewhat academic). I guess if you
> > deny an association with Jigsaw, I can understand your inability to
> > respond to concrete questions with concrete references.
> 
> I can't help but get the impression that you have misunderstood the
> purpose of this mailing list and the nature of the discussions here;
> you appear to be putting the cart before the horses. The experts on
> this list (who graciously allow us to review and comment on their work
> as it is in progress) are working to define what JSR 294 shall be. By
> necessity they cannot reference the completed specification because
> that is the end goal of their work here!

I would suggest that the experts to whom you speak may which to get
their house in order.  The JSR 294 process was formed more than three
years ago (09 May, 2006).  Since formation a lot of things have happened
which you have neglected to reference. The 277 process was born and died
(and lets not get into the reasons for that).  But something positive
came out of that - a implementation project called Jigsaw.  I would
suggest to you that it is not a case of me putting a cart before a
horse, but instead, it is much more of a case of the horses making a
vain attempt to keep up with a smart cart.


> > At least the Jigsaw project is pumping out details of language changes,
> > VM implications, code, APIs and a sample implementation.  One could
> > suggest that that the Jigsaw Project is a more credible reference point.
> > In reality (and I think you, Peter, and Richard would agree), Jigsaw is
> > delivering.  In the meantime one could argue that people are wasting
> > bandwidth with discussions that are not grounded on specifications.
> 
> Presumably you feel that all JSRs should begin with a complete
> specification and work backwards from that point, and that any
> discussions prior to the production of that specification are a waste
> of bandwidth?

No, not at all.

I do expect that JSRs provide timely deliverables.  I do expect
'experts' to behave as such and demonstrate real expertise. I do expect
special interest to be obstructive.  I do not expect nor am I willing to
tolerate obstruction in the form of ineffectual flag-waving.  I do not
expect a failure of experts to address basically simple concrete
questions.  Perhaps my expectations are too high - thing is - that's
your problem.  I expect result - good or bad.  What I am seeing is a
collection of wannabe experts circulating around external event without
the collective testicular fortitude to come up with a position.

Now perhaps you create an cunning reply detailing your baseline - a
bunch of URLs, detailing your position? Give me your baseline and I will
give you some credit.  Until then, you remain on the sideline.

> > I don't think I'm ready to drop Jigsaw spec. from the equation - in
> > fact, I would suggest that you and the other members of the expert group
> > should deliver a baseline specification ASAP. You could publish it, and
> > then address issues relative to that baseline. At then end of the day, I
> > think it is important to ground our discussions on qualified proposal
> > and tangible acronyms.  I sure you agree.
> 
> Jigsaw is not a reference for JSR 294, it is a potential future
> implementation of JSR 294. It may be further advanced than the
> discussion here but that is irrelevant. OSGi is also a potential
> future implementation, and it is even further advanced, being in
> production already in millions of devices and products, yet I do not
> see you insisting on any reference to the OSGi specification as a
> baseline.

Oh please - get a grip.  Jigsaw is certainly further advanced than the
so called expert group.  On this we are in complete agreement.  As to
OSGI, this is legacy and you and I both know that this is background
technology.  Once you break appart OSGI's notion of modules from OSGI as
a service delivery platform - you quickly realise that it's not rocket
science.  Isn't it time to walk away from smoke and mirrors?

> It is proper for the requirements of Jigsaw to influence these
> discussions, 

Damn right it is. And it is equally proper that the 294 Expert Group
provide appropriate recognition, support, and engagement, and challenges
where required.  It is also completely inappropriate for members of the
so called expert group to call for the closure of 294.  Where were you
on that event, where where you colleges, where was accountability?


> and Alex is doing a fine job in communicating those
> requirements to the group. 

No problem with Alex - although sometime I wish he would call a dork a
dork (but if I were in his position I would working hard on constraint,
self control, etc.). I love Alex!

> It is also proper for the requirements of
> OSGi to influence discussions, and BJ, Peter and Richard are doing a
> fine job in that regard also. 

Sorry - I disagree.  BJ has failed to detail his concerns, Peter has
been flip flopping and waving flags.  Richard has failed to provide a
compelling argument and all of this has a strange familiarity.  

> It is NOT proper to short-circuit the
> entire JSR process and arbitrarily choose one incomplete
> implementation as the baseline of a future specification.

Hang on - you started back in 09 May, 2006 and you have the balls to
write the word 'short-circuit'? I'm really impressed, I really am.  OK,
let's take a step back and look at the big picture.

The JSR 294 project [1] aims to deliver a specification that will
establish a definition of modularity for the Java language.  Initially
this project was deployed with a close connection to the JSR 277
process. The 277 process [2] was suspended.  The 294 process started out
(09 May, 2006) with a open policy, publishing intermediate results,
however it has been a while since something of substance has been put up
as a definitive position. 

In October last year, the 277 process went into permanent hibernation.
We are now almost one year after that date.  Almost the anniversary.
While 277 was thrown to the sunset, a new project named Jigsaw was
established. Over the last year Jigsaw has published an initial language
change specifications [3], class file changes [4], VM changes [5], notes
on the compiler architecture [6], an API under the java.lang.module
namespace and an initial implementation [7] (not to mention a slide show
detailing the thought process leading to compile-time mapping of module
source information [8]).

Earlier today we had an email that attempted to position Jigsaw as
something akin to a loose cannon in that Jigsaw is just an OpenJDK
project that may influence things relative to the 294 process.

I would suggest that that your position is irresponsible and naive. In
the absence of a concrete proposal by the 294 Expert Group, I would
suggest that the Jigsaw documentation is the standing reference. If this
is not the case, I strongly recommend that the members of the expert
group state this as a formal opinion.  If the expert group cannot
endorse Jigsaw, then an obligation rests with expert group members to at
least publish their baseline within an reasonable timeframe.  If the
expert group cannot see an a reasonable avenue to achieve this goal, I
would suggest that an expert refresh may be in order.

Cheers, Steve.

[1] http://jcp.org/en/jsr/detail?id=294
[2] http://jcp.org/en/jsr/detail?id=277
[3] http://openjdk.java.net/projects/jigsaw/doc/language.html
[4] http://openjdk.java.net/projects/jigsaw/doc/classfile.html
[5] http://openjdk.java.net/projects/jigsaw/doc/vm.html
[6] http://openjdk.java.net/projects/jigsaw/doc/javac.html
[7] http://cr.openjdk.java.net/~mr/jigsaw/api/
[8] http://openjdk.java.net/projects/jigsaw/doc/ModulesAndJavac.pdf


> 
> Kind regards,
> Neil
> 
> >
> > Cheers, Steve.
> >
> > p.s. Still waiting for an answer on the single classloader question and
> > that other question about implementation specifics data in the
> > module-info data related to the Jigsaw implementation.
> >
> > Cheers, Steve.
> >
> >
> > --
> > Stephen J. McConnell
> > mailto:mcconnell at dpml.net
> > http://www.dpml.net
> > mobile: +61 4 5800 3980
> >
> >
> 
>