[ofa-general] New proposal for memory management

[Jeff Squyres]

On Apr 30, 2009, at 1:51 PM, Barrett, Brian W wrote:

> All we want is *SOMETHING* we can do that we know is safe.
>

This is the most important thing.  +1

> The registration cache seems like the safest,
>

+100

> but the notifier would be better than where we are today if we can  
> get the important race conditions out of it.
>

+0.1.  Yes, it would work, and I agree that it would be marginally  
better than what we have today.  IMHO, it seems like fixing this Right  
would be better than "let's do a possibly simpler but sub-optimal  
solution."

> The current state of the field is unacceptably stupid and I'm actually
> amazed there's any resistance to fixing the problem.
>

I'm a little amazed that it's gone this long without being fixed (I  
know I spoke about this exact issue at Sonoma 3 years ago!).

-- 
Jeff Squyres
Cisco Systems