Re: [picocontainer-dev] Re: [picocontainer-scm] [3768] java/2.x/trunk/pico/container/src/test/org/picocontainer/behaviors: reentrant lock stuff - alternative to synchronized

peter royal wrote:

> On Jul 28, 2007, at 4:59 PM, Paul Hammant wrote:
>> Jörg mentioned it some months back. Suggesting we could switch to
>> it from synchronized.  The IBM article suggests its better for now,
>> but not forever.
>>
>> Rekindling that discussion.
>>
>> Should we roll this into Caching or is it best to have things
>> separate?
> 
> I'm not sure the use-case of having it separate.

Ahh no, not again, thought we were over it ;-)
Look into the archives for the discussion 2 months ago.

> And if it is separate, even less sure of the value of having two
> implementations of the same thing. Sounds like it'll be begging for
> confusion from users.

That's true. We should take the one that's working best. The reentrant
locking will avoid that you pay for synchronization for reading. However, a
small benchmark should reveal it. Have a look at XStream's benchmark
package ... it is not tied to XStream :)

- Jörg

BTW: I've started for XStream to keep such little benchmarks for separate
aspects of the framework. I normally run these tests against the last
release and the head revision - it is quite helpful sometimes.


---------------------------------------------------------------------
To unsubscribe from this list please visit:

    http://xircles.codehaus.org/manage_email