Updating doesn't give us any advantages for the core framework,
Java 5 is minimum version requirement and newer Java versions work as
well. Or do you plan to refactor something which would depend on Java 6?

It is a different story for sandbox/addon projects, there we could make this decision
independent of the core framework.

Anyway I wouldn't mind to switch to Java 6.

It seems more interesting to update the eclipse version, since it ages
quicker than Java, and there are some features here and there we cannot
use in the Cas Editor or must use now deprecated APIs.

Jörn

On 11/16/11 9:45 PM, Tommaso Teofili wrote:
Hi all,

looking at the roadmap towards 2.4.0 release of UIMA SDK I was wondering
why not migrate to Java 6 and consequently drop Java 5 support.
The 2.3.1 version is known to be stable enough to keep it as the last Java
5 enabled version in my opinion to allow backward compatibility.

This way we may also open up to integrate easier with a number of other
open source projects which adopt Java 6.
Just as an example, some months ago I was working on the RDF CAS Consumer
using Clerezza but then (my fault) I realized it was not possible to use it
on Java 5 as Clerezza uses Java 6 however this alone wouldn't be a strong
enough reason.
Other much more important reasons, in my opinion, lie in bugs in Java 5
which have been fixed only in Java 6, see this one from Lucene as an
example:
https://issues.apache.org/jira/browse/LUCENE-3235

Consider also that some bugs have been resolved in minor versions of Java 6
so this wouldn't secure things 100% from previous bugs, but I think that
doing a minor version upgrade (usually to at least 1.6.0_18) is trivial
while changing major versions could be an issue with a big project.
By the way, it seems Lucene 3.x is going to drop Java 5 support thus this
would block any further update to Lucas/Solrcas (Lucene/Solr 3.4 still has
Java 5 but 3.5/3.6+ will probably migrate to Java 6).

What do you think?
Tommaso


Reply via email to