Thanks Andy.

I was going to ask about this non-incubating release, but this answers it 
:-)

I have been testing a 2.7.1 snapshot of earlier in the week and have found 
no functional problems so far. Query performance with 50+ threads is still 
somewhat weak, but I am going to compare the same suite in a 2.7.1. 
without transactions and our 2.6.3 build. I can send you the numbers once 
I have them. Moreover, one of our client teams are conducting a very large 
experiment as well (1 million+ resources/graphs) and 200+ users/threads. 
As expected memory consumption is higher with Tx, but still manageable. 
What did surprise us is that the physical store nearly doubled from 35Gb 
to about 70Gb.

As for the issues you write below:

1) single source release is nice for our open-source approvers
2) keeping fuseki separate from apache-jena is also great. We typically 
only need iri-core-arq-tdb :-)
3) no opinion on 3/ and 4/
4) LARQ: so far, we have not been using it, but we are thinking of 
starting with that, so having it released with 2.7.1 would be nice. If 
this does not happen, then, it would help if there is a clear indication 
with what release a given component would be compatible (that request 
would apply to any component which is not released with the "main" pack

thanks

Simon




From:
Andy Seaborne <[email protected]>
To:
[email protected]
Date:
05/25/2012 09:30 AM
Subject:
Next apache-jena release



It would be nice to make a release now we are out of incubation.

I think doing the least to make the release happen is the right next 
step but I can be persuaded that doing more on integration of components 
is worth waiting for.

Starting proposal:

1/ Single source release, not one per module.

2/ Two downloads:
      apache-jena being iri-core-arq-tdb
      jena-fuseki being Fuseki
      (==> switch off assemblies for ARQ and TDB for the release)

3/ Separate versions numbers for component, numbering as we have in svn 
at the moment
    . we haven't experimented with an uber jar yet
      shouldn't be any problems other than verbosity of the plugin
    . separate jars mean we can release updates
      for example, fine-tuning for final SPARQL stuff.

4/ The parent is not a separate release.
    If we do a single build, I can't think of a reason
    for a separate parent but I may have missed something.

Thoughts?

Any JIRA that you feel a strong need to complete and get in the next 
release?

Paolo - what about LARQ?  If it gets placed in the release, would that 
hamper getting incremental update released soon?  Or do as a parallel 
release (i.e. not in apache-jena but part of the build) to get the 
version moved on to non-incubating.

Next step:

I will set up a Jenkins job to build (2.7.1-snapshot) apache-jena as a 
preliminary step and then disable  the explicit snapshot builds of 
iri-core-arq-tdb.

We can inspect the binary it builds ... and it will at least mean no 
ordering problems with snapshot builds.

                 Andy

PS And no more icu4j!



Reply via email to