> IOW, whatever is specified in the POM file of individual projects is
> likely to be overridden anyway and the best the could be done
> is to avoid known gaping incompatibility holes (of which I'm not
> aware of too many).

Fair enough, but Guava is a particular problem. Map eviction listeners
completely change somewhere between r09 and r11. It could
theoretically be possible to handle that with reflection but that
would be a lot of reflection, in performance sensitive code paths, and
likely to fail only at runtime/in production if there is a mistake or
another change. So all of Hadoop should get on the same page with
this. Sadly I see Giraph just moved to 12 while everyone else is on
11.

On Sat, Jun 30, 2012 at 7:17 PM, Roman Shaposhnik <[email protected]> wrote:
> On Sat, Jun 30, 2012 at 5:24 PM, lars hofhansl <[email protected]> wrote:
>> So will Guava 11.0.2 in hbase 0.94 be a problem? Should we not update Guava 
>> (at least not until 0.96)?
>
> In general, I don't think there's good answer to how dependencies
> are supposed to be managed wrt. version numbers. After all, all
> of the Hadoop ecosystem projects have to deal with the fact that
> there's at least half a dozen different branches of Hadoop they have
> to work with. It is virtually guaranteed that whatever you do you will
> end up incompatible with one of those builds and most likely the
> version of a dependency that wins will be the one that Hadoop is using.
>
> IOW, whatever is specified in the POM file of individual projects is
> likely to be overridden anyway and the best the could be done
> is to avoid known gaping incompatibility holes (of which I'm not
> aware of too many).
>
> Thanks,
> Roman.



-- 
Best regards,

   - Andy

Problems worthy of attack prove their worth by hitting back. - Piet
Hein (via Tom White)

Reply via email to