Subj : Re: Formal Mutex Semantics
To   : comp.programming.threads
From : Maciej Sobczak
Date : Tue Aug 16 2005 10:04 pm

Joe Seigh wrote:

> Current thread usage patterns would break, e.g. initializing an object
> and queuing it onto a synchronized queue.  You'd have to either iniitialize
> the object while holding the queue lock or copy it while holding the queue
> lock.  Either way would increase the lock hold time and the probability of
> contention.

Indeed.
What's worse, some objects have inherent shallow-copy semantics (like 
descriptors and handles to system and external resources), which means 
that this model would not be very comfortable to use.

Thanks for the explanation.


-- 
Maciej Sobczak : http://www.msobczak.com/
Programming    : http://www.msobczak.com/prog/

.