Re: Standard Library problems w/multi-threading

From: williamkempf_at_[hidden]
Date: 2001-09-10 22:15:53

--- In boost_at_y..., "Jeff Garland" <jeff_at_c...> wrote:
> 
> > After attempting to implement a Date/Time class several times and 
never
> > really getting a satisfactory version I would love to see a Boost 
library
> > with an eye towards eventual submission to the standards 
committee. If
> > anyone tries this I hope they will avoid the pitfalls inherit in 
the old C
> > time functions. In particular it should be possible to represent 
any
> > reasonable date and not be limited to a scale represented by the 
storage of
> > seconds since 1/1/1970.
> >
> > Ken Shaw
> 
> Take a look at this message:
> 
> http://groups.yahoo.com/group/boost/message/10120
> 
> I believe I should have this ready for public discussion in the 
next month or so
> :-)  In GDTL the "clock source" is a component that could be 
implemented with
> the C functions, a network time source, or other implementation.  A 
date or time
> object can be constructed from the clock source, but the date class 
is not
> restricted to the date epoch used by the clock source (it can be 
larger).
> However, the default implementation for the clock is the C time 
functions, so
> this threading issue will need to be examined.

I'd really hope that we'd first implement a full xtime 
implementation, since this is likely to be standardized on by C in 
the near future.  This type has addressed most (all?) of the 
shortcomings of the current C/C++ date/time handling routines.  Then 
a boost date/time class and duration can be built based on top of 
this specification, giving us a more OO interface to a (soon to be) 
common C type.

Again, I love the idea of being able to code something like:

cond.wait(boost::calendar() + duration::from_seconds(1));

Bill Kempf