RE: [boost] Time interfaces threads/file (was Design for a file s ystem library)

From: Darryl Green (green_at_[hidden])
Date: 2002-02-27 00:34:07

> -----Original Message-----
> From: Jeff Garland [mailto:jeff_at_[hidden]]
> > One thing I don't see explicit mentioned here is the 
> concept of a monotonic
> > clock as per IEEE Std 1003.1-2001. Is a clock type which 
> deals with elapsed
> > time and is unaffected by changes to the system time supported?
> 
> Yes agreed.  Seems like this would be mostly an issue for the 
> threading library.
> 

It is a big issue for the threading lib - which imho should allow the use of
both monotonic (need a better word , I don't mean TAI) and "time of day"
time types for specifying timeouts. I don't know how much work is involved
in a portable implementation of Boost.threads that allows both the UTC clock
type currently supported and a monotonic clock type. It probably isn't a big
issue for GDTL. Is the time type I'm talking about simply a clock system
with the epoch being some arbitrary time at which we stated caring (in
practice, this is probably system start time, but for threads could be as
late as program start time) that is monotonic, and uses a system dependent
clock that is not affected by any adjustment to "system time"?