2012/6/21 Thomas Jungblut <[email protected]> > We can make option 3, but with a new submodule of Hama instead of github. >
the above, with IO stuff in a new submodule, is by fair my preferred one. Having things in github wouldn't make much sense if it's something the Hama PMC and users agree to have here. Regards, Tommaso > > But 3,2,1 is mine. > > 2012/6/21 Praveen Sripati <[email protected]> > > > What is the approach Hama should take? > > > > *Option 1* - Put hama input/output formats in Hama > > Dependencies have to be included in Hama which will make it heavy. But, > > nice thing is that all the Hama Input/Output formats will be in one > > location. Also, maintenance is a continuous process with the version > > upgrades and the API changes in the related projects. > > > > *Option 2* - Put hama input/output formats in cassandra, accumulo, hbase, > > ..., etc > > A lot of Hama code will creep into other projects and also we need to get > > approval from other projects also. Some of the projects are not open > source > > and it might not be possible to work with them to include the code. > > > > *Option 3* - Put hama input/output formats on something like github > > Users can pick what they want and build them. Hama is also light. But, > > this might be something new to the Apache way. > > > > Here is my preference 3, 1, 2. > > > > Please let me know your preferences to reach a consensus. > > > > Thanks, > > Praveen > > > > > > -- > Thomas Jungblut > Berlin <[email protected]> >
