I also am +1 for option 2. #3 is my least favorite, because of the download time for some of the models, both for cases like Steve mentioned but also for cases of wanting to check out a fresh copy of the code and not wanting to wait to check out the models again
-- James > -----Original Message----- > From: ctakes-dev-return-1378-Masanz.James=mayo....@incubator.apache.org > [mailto:ctakes-dev-return-1378-Masanz.James=mayo....@incubator.apache.org] > On Behalf Of Steven Bethard > Sent: Friday, March 15, 2013 1:06 PM > To: [email protected] > Subject: Re: [DISCUSS] Where should cTAKES models live? > > On Mar 15, 2013, at 4:39 PM, "Chen, Pei" <[email protected]> > wrote: > > So the question is: What should we do with the model files? Some > options include: > > > > 1) Leave them in SourceForge/Maven Central. Maven can download and > include them in the convenience binaries in the ctakes-distribution > project. Something we did quickly for 3.0, but needs to be improved if we > go with this approach. For example: [2] > > > > 2) Leave them in the ASF repo, but separate modules/projects. > > > > 3) Keep them in the same respective ASF modules under > /src/main/resources > > > > I think it's nice to keep these fairly large (~1GB) and static resource > files separate from the source code (Either option 1 or 2). Also, option > 1 will require a little more work by the committers/release managers but > will definitely avoid any licensing issues/concerns. > > I'd definitely vote for (2). That makes releases much easier than if you > have to coordinate between the ASF and Sourceforge repositories, but also > allows people to depend on the code in a module without also pulling in > all the models as well. (This would make a lot of sense even now, for > example, in ctakes-temporal which depends on ctakes-relation-extractor > only for the relation extraction framework and not for the location_of and > degree_of models.) > > Steve
