On Jul 23, 2012, at 11:05 AM, Mattmann, Chris A (388J) wrote:
> On Jul 23, 2012, at 9:38 AM, Steven Bethard wrote:
> 
>> On Jul 23, 2012, at 9:53 AM, Mattmann, Chris A (388J) wrote:
>>> 2. Getting the 2.6 codebase for cTAKES in SVN ASAP and seriously 
>>> considering making the 2.6 release *here* at the ASF. 
>>> What is the problem with doing that? Are you concerned that it will add 
>>> extra time?
>> 
>> I believe that's at least one of the concerns - wouldn't we at least need to 
>> change all package names to org.apache.ctakes before the first Apache 
>> release?
> 
> That's not *absolutely* required, and it's more preferable to at least 
> release at the ASF keeping the old
> package names, than to release outside.
> 
> Besides that though, isn't changing the package names something that could be 
> easily scripted? We did
> so for Apache OODT within a matter of a few days during Incubation with some 
> Perl and Python scripting.

Changing them is not trivial because of UIMA's use of XML descriptors, which 
makes tracking down all the places where you might need to change a package 
some extra work. I agree it could be scripted, but the additional work is in 
getting all the tests running again and other quality control measures that are 
necessary before a release.

Another reason not to change the package names before the 2.6 release is that 
there are other package name changes that probably should be made, and it would 
be best to make all package name changes at the same time we change to 
org.apache.ctakes. (Otherwise we have a more painful upgrading process for the 
users.)

But since the package name is not required, maybe a way forward would be to:

(1) Just get cTAKES, as is, into the Apache repository, and make the 2.6 
release exactly as planned (no code changes at all) from the Apache repository.

(2) Immediately after the 2.6 release, change the package names to 
org.apache.ctakes, and use that for the 2.7 and subsequent releases.

Steve

Reply via email to