[I renamed the discussion thread to separate it from the vote thread.]
I agree with Michael that the first and most obvious benefit is
merging cayenne-jdk1.4-unpublished and cayenne-jdk1.5-unpublished in
a single module and getting rid of 1.4 workspace.
Separate from that, IIRC Ari have already voiced concerns about 3.0-
final being too far in the future. Why would that be a problem? We
are releasing high quality milestones that people can use in
production, but beyond that I feel like we don't have "completeness"
of the new feature set that would warrant a final release. I am open
to discussion on that, but so far I don't see a reason to rush.
Andrus
On Aug 14, 2007, at 8:39 AM, Michael Gentry wrote:
"Will moving to JDK 5 be a change in label only, or will we actually
go in and implement generics throughout all the classes, possibly
requiring some API changes along the way to do it right?"
I'm thinking it'll be a label-change only (well, except the JPA stuff
does indeed require Java 5), but would allow us to start integrating
Java 5 features into the main code baseline (which would make Cayenne
3.0 incompatible with Java 1.4). I don't imagine any of us are
wanting to go refactor working code at the moment, but we can do ad
hoc changes as parts of the code are being worked on.
"Will this then add 3-6 months to the release of 3.0..."
Right now, Cayenne is broken into 2 Eclipse projects -- one for Java
1.4 and one for Java 1.5. To do development between the two, you
should switch Eclipse workspaces (or run Eclipse twice), which is a
pain. And I did mean switch workspaces, not switch projects. I would
think having the entire Cayenne 3.0 release codified under a single
Eclipse project would simplify and improve the development effort, not
cause real delays.
/dev/mrg
PS. I'm +1 for Cayenne 3.0 to require Java 1.5.
On 8/13/07, Aristedes Maniatis <[EMAIL PROTECTED]> wrote:
On 14/08/2007, at 4:11 AM, Kevin Menard wrote:
Since this is a bit of a big change for us, and Ari had expressed
some
concern, I'd like to just take a vote to ensure this is the
direction we are
heading.
I certainly see the benefit in the longer term, but I wonder what the
purpose of this change is right now half way through the 3.0
development process. Personally I'd like to see 3.0 out as soon as
possible, since it has been a long time since the last stable new
feature release (1.2) and I think releases help keep up the perceived
momentum of the project.
Is the plan for 3.0 to not release until it has full JPA compliance?
If so, a release this year seems unlikely.
Will moving to JDK 5 be a change in label only, or will we actually
go in and implement generics throughout all the classes, possibly
requiring some API changes along the way to do it right? Will this
then add 3-6 months to the release of 3.0 while the changes are
ironed out? Are there other JDK 5 features we desperately want to
use? I know that the Swing improvements could make the modeler nicer,
but that too requires a whole bunch of work. I've just done a lot of
work putting generics into my major project, and it isn't always as
simple as pressing the 'add generics to this class...' button in
Eclipse.
* generics
* swing improvements
* nicer for loop (but very minor functional change or speed
improvement)
* other little things
So my hesitation is to do with feature creep. If moving to 1.5 adds
to the release schedule considerably, then I'm -1.
Sorry to sidetrack this vote with questions, but I'd like to be clear
about the benefits/costs of this decision.
Ari Maniatis
-------------------------->
Aristedes Maniatis
phone +61 2 9660 9700
PGP fingerprint 08 57 20 4B 80 69 59 E2 A9 BF 2D 48 C2 20 0C C8
