[LEAPSECS] LEAPSECS Digest, Vol 82, Issue 6

[Joseph Gwinn]

On Sun, 11 Aug 2013 14:21:41 -0400, Greg Hennessy wrote:

> 

>> The POSIX standard in fact forbids it, precisely to escape the time

>> wars.

> 

> If the guys writing POSIX think that forbidding POSIX from matching

> reality means no time wars, I think they have another think coming.


No, we already had the time war.  Twice.  It's a very long story, 
documented in the TimeNuts archive, which in turn points to the noisy 
debates in the POSIX world.



>> However, POSIX allows an implementation (read, operating system) to

>> define as many named clocks as desired.

> 

> I have never head of this before. How does one define a (for example)

> UTC clock in a POSIX system?


As a user, you do not.  As a platform developer such as Linux kernel 
developer, you can.

Here is some debate about two named clocks in POSIX/Linux: 
<http://stackoverflow.com/questions/3523442/difference-between-clock-realtime-and-clock-monotonic>

The general interface used by these two named clocks can be extended by 
a kernel developer.

At this point, people usually ask why this could not also be provided 
to the user.  The problem is that one cannot just whistle a new kind of 
clock up.  The hardware and kernel must support it, and these things 
cannot be provided by a user.

Joe Gwinn