[JSR308] JSR 308 documents and tools released
Tom.Ball at Sun.COM
Sat Jul 7 00:46:19 EDT 2007
Michael Ernst wrote:
> Thanks for your message, which reiterates points you have made before.
>> there is no BNF
> It's evident that you are extremely anxious about a formal BNF, and one is
> needed before the proposal is finalized. However, it's frankly not as
> important as other issues on our list, and it doesn't seem to be causing
> confusion for anyone but you.
That's not true -- I'm confused, too, but since Neal's arguments have
nailed all of my concerns I haven't spoken up in support of them. Other
engineers I've discussed JSR 308 with at Sun are also anxious about it,
and not just by its lack of a formal BNF. Major changes to the Java
language require a higher level of detailed specification than most
products require -- the Java community hasn't always lived up to that
standard (and Sun has certainly contributed its share of mistakes), but
each time we leave things unspecified or specified incorrectly, those
mistakes have come back to haunt all of us.
>> - The implementation techniques are inconsistent with javac's design
> We are aware that the implementation is not perfect; it's a prototype
> intended to let people get experience with the system. I encourage you to
> do so; you might find you like it better once you actually try it.
Maybe, but an implementation that isn't maintainable shouldn't be part
of the JSR's overall deliverables. Neal wasn't raising nits in a
prototype, but instead describing basic design decisions in the
prototype that can't be sustained in *any* production-quality compiler.
Please make his concerns a high priority when working on improvements
to that design.
> How about sending a patch that fixes these things? (Even if you don't like
> the syntax, you could easily make the fixes that don't have to do with the
> parser; but we would gladly accept parser fixes as well.) That would be a
> great way for you to make a productive contribution to JSR 308.
Neal has already made such a contribution, by clearly detailing specific
issues with the current specification and implementation drafts. I am
concerned that his concerns are not taken as seriously as they should
be, as your comment implies.
More information about the JSR308