From: Paul Baxter (pauljbaxter_at_[hidden])
Date: 2005-04-24 08:24:27
< lots of good suggestions snipped >
I agree that threading needs to find its way into the standard and hope
enough progress is made soon to make this a reality. I hope many aspects
from the shmem library and others are considered to enrich the basic
functions currently offered by boost threads. This would improve the utility
of a 'threads' submission in my opinion.
I also think that thread priority and scheduling need to be implemented in
some form. I appreciate that in certain OSes this isn't possible, but if
such priorities/scheduling types were hints which an implementation were
free to ignore then why couldn't these feature as say optional constructor
parameters or in a member function for adjusting priority/scheduling type.
I couldn't find the mailing list discussion (that I expect did happen)
regarding the reasons not to include priority etc.in the interface, but
would be interested if their inclusion may be possible in future revisions
of boost thread.
I also find the notion of creating a new thread and monitoring an existing
thread being represented within the same class (with different constructors)
as being slightly non-intuitive. This may be one of many areas where boost
threads could benefit from a more tutorial-based set of documents rather
than a purist model/concept set of docs.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk