Proposal for discussion - remove all Windows-port cruft

[Hal Murray]

ghane0 at gmail.com said:
> In any case, while I can imagine the client-side and monitoring code running
> through the shim, how likely is it that Gary and Hal (and Dr Mills) PLL code
> would survive being machine-translated to a non-POSIX kernel?

There are 2 interesting parts to the kernel.  The important one is adjusting 
the clock rate, the drift.  With that you can run a good NTP client which can 
also be a server.

The other part is the PPS stuff.  The critical part is to get a time stamp on 
the PPS signal.  The kernel PLL is not required.

----------

There are 2 styles for making code run on various environments.  One is to 
sprinkle #ifdef and blocks of code inline.  The other is to make an API that 
covers what you need and then write an implementation of that API for each 
environment you are interested in.

In this context, I'm willing to see how well POSIX works for that API.  It 
would be interesting to see how much of our code runs on cygwin or something 
similar.  If the only problem is adjusting the clock, we can probably figure 
out how to do that and write a wrapper.

-------

The NetBSD kernel for the Raspberry Pi doesn't have the critical kernel call 
implemented.  I haven't investigated.   They ship a version of ntp classic 
that works.



-- 
These are my opinions.  I hate spam.