[Ns-developers] SingleEvent and EventCollector

Mathieu Lacage mathieu.lacage at sophia.inria.fr
Mon Jul 30 04:05:52 PDT 2007


On Mon, 2007-07-30 at 11:57 +0100, Gustavo Carneiro wrote:
>         
>         yes, it will keep the EventImpl alive after the event has
>         expired. That
>         issue notwithstanding, I think that it makes sense to make the
>         cleanup
>         heuristic be able to decrease m_nextCleanupSize.
> 
> Sure, but anyway, that is an implementation/optimization matter and
> doesn't change the API.

Indeed. So, maybe, if you want to avoid having to deal with the
benchmark, you can just remove the cleanup heuristic and
m_nextCleanupSize altogether and we can work on it when we have a better
handle on usage patterns at which point our benchmark will be easy to
build :)
> 
>         >  lots of EventId's.  And if it keeps the EventImpl after
>         expiration,
>         > why you need it so?
>         
>         Because, if you looked at bug 58, the proposed fix is to
>         refcount
>         EventImpl so, if it is refcounted, you have to keep it alive
>         until all 
>         references disappear. Now, it is quite possible fix the bug in
>         a
>         different way which does not require refcounting: it would
>         require using
>         a "reference list" but refcounting was the simplest way to
>         implement 
>         this. If you could show a use-case which really shows a
>         problem because
>         you keep too many EventIds around, that would be a good reason
>         to switch
>         to the more complicated solution.
> 
> Yes, you're right.  There's no need to create a more complex solution
> (anything more complex would only degrade performance anyway). 

exactly.

Mathieu

-- 


More information about the Ns-developers mailing list