Re: [udig-devel] Java 1.5

[Chris Holmes <cholmes@xxxxxxxxxxxxx>]

On Thu, 2 Dec 2004, James Macgill wrote:

> At 02:51 PM 12/2/2004, Jody Garnett wrote:
> >David Zwiers wrote:
> >
> >>Hello all,
> >>It appears that we should once again talk about the possibility of
> >>transitioning to Java 1.5.
> >>Advantage:
> >>    - Clarity
> >>
> >>Disadvantages
> >>    - Mac Support is limited (but so is uDig's current Mac support)
> >>
> >>
> >>Currently much of Java1.5 is already used, but only as part of the
> >>javadoc comments.
> >>
> >>Any thoughts?
> >>
> >Mac support for java 5 is expected in the spring.
> >
> >Generics, Enums and VarArgs are the main advantages for clarity - these 
> >are strong advantages given that we are defining API.
> >
> >+1 for me, although I am frustrated with mac development.
> >
> >Chris/James did you have any thoughts on this? I know you are in touch 
> >with a few other open-source groups.
> >
> >Jody
> 
> Remember to make a distinction between Java 5 runtime and Java 5 compile 
> time.  iirc many enchantments, inc generics will run under Java 1.4 but 
> will not compile under it.
> 
> Given that the pool of Mac 'developers' is likely to be small (as opposed 
> to users) you may be able to use a subset of Java 5 language enhancements 
> (but not API enchancements) without impacting portablity.
> 
> Not that I've tested the above theory!
> 
> James 
> 
First off, I'm supporting 1.4 in GeoServer for a long, long time.  There 
are two reasons for this.  The first is that as a server project, the real 
users will be people who run it on servers that are already set up and 
configured, and on those types of machines people are _much_ more hesitant 
to upgrade.  They don't want to mess with something that already works, 
especially when many people are using it.  Related to that J2EE evolves 
behind J2SE, so it's only just now getting to 1.4 I think.  So that really 
doesn't apply to uDig, as it's a client project, and people can easily 
upgrade java, indeed the best route is for you to just include, so people 
don't have to deal with it at all.  The second applies more, and that's 
being _truly_ open source.  That is to say by going with the bleeding edge 
you are dependant on Sun.  Java's almost but not quite open source nature 
makes it so other JVMs, especially on open source operating systems like 
FreeBSD, lag behind.  This may change a bit as Sun seems to be making 
moves to make Java more open source, but even still it'll take time to 
port other operating systems, many are just reaching 1.4 now.  The GNU 
classpath project (I think that's what it's called, a gnu jvm) is one that 
hasn't yet done 1.4.  So in going with 1.5 you cut out a small, but 
potentially significant, population.  I say significant because many times 
the people who are strict about open source are very willing to contribute 
to new open source projects that they like, and are generally very good.  
This is sort of the flip of James's argument about mac developers vs. 
users, because most of your users of more obscure open source operating 
systems (or hardcore gnu fanatics), are probably good developers (yes, I 
know I'm making blanket statements here).

So the first is not an issue for you, the second depends on how much you 
care about being _truly_ open source.  I personally would stick with 1.4, 
but I also haven't examined 1.5 and the benefits it offers, indeed am kind 
of willfully keeping away, since I'm committed to 1.4 for a long time.  
The one thing this does bring up is the fact that if/when you roll changes 
back into geotools you'll have to back port to java 1.4.  Which may not be 
a big deal, I really don't know, but it's there.  And I'm also a bit 
against it, just because I know the day to try to upgrade geotools will 
come sooner if we have most developers working in 1.5, and I am very 
against upgrading geotools.

As for using language enhancements and not API enhancements, I'd say 
that's hard to do, as you'd have to willfully just look at 1.4 javadocs 
(and maybe some how set your function completion in ide back).  It's way 
too easy to just put in a nice new method that is in Java 1.5.  Mostly 
because the new ones make a lot of sense, you wonder what people did 
before that.  This happened just the other day with a contribution I got 
to GeoServer, in the Matcher class.  

So that's my take.  But I can understand if you'd go to 1.5, even though I
would recommend against it, as I admit my reasons are not especially
compelling.

best regards,

Chris