The best practice here is to develop all software which should go into
cTAKES here at Apache,
doing that has a couple of advantages:
- Nobody is excluded from participating (contributing code, discussion
on the mailing list or jira, etc.)
- There are no IP issues
- All the history is in subversion
I suggest to create a sandbox area or something else to work on things
which are not yet
ready for cTAKES trunk.
Jörn
On 10/10/2012 05:59 PM, Steven Bethard wrote:
As part of the THYME project [1], we're developing cTAKES modules for
identifying events, times and temporal relations. Where should these go? I see
a few possibilities:
(1) We add a ctakes-temporal module to ctakes trunk. This is good because it's
where the final module will go, it makes it easy to compile with any
cross-project dependencies, and anyone makes any changes that break the
ctakes-temporal code, they'll be immediately obvious with compile errors. The
downside might be that if it's not 100% ready for release by the time the
3.0.0-incubating release rolls around, we need to find a way to exclude it from
that release.
(2) We host it somewhere else and move it into ctakes when it's 100% ready for
release. We're already doing this now, at
https://github.com/bethard/ctakes-temporal, but we could also put it in the
cTAKES sandbox. The downsides are that it's harder to get the cross-project
dependencies to work (especially on the command line with Maven and SNAPSHOT
versions), that breaking changes in cTAKES will only be visible to people who
have separately checked out ctakes-temporal, etc.
What do people think?
And in general, what should be the process for introducing new cTAKES modules?
Steve
[1] http://clear.colorado.edu/compsem/index.php?page=endendsystems&sub=temporal
