[ntp:hackers] Reference Clock code for Windows - How can I contribute?

Danny Mayer mayer at ntp.org
Sun Apr 27 03:01:19 UTC 2008


William Horner wrote:
> Danny,
>
> I am assuming that you are interested in reference clock driver code for 
> Windows?  I am also interested in seeing such drivers become more 
> mainstream because NTP is such a good piece of software.
> 

Of course we are interested.

> This is a note to ask about how I might be able to contribute modified 
> Windows reference clock code back to the NTP project?  Are you accepting 
> such contributions at the moment, or are you getting too close to a 
> stable release?

See Harlan's response on how to do that.

> I have managed to get the ARCRON reference clock driver working in 
> Windows.  Mainly because I have got two such clocks and they are 
> surprisingly common in the UK (under various different manufacturer 
> names) and although not very precise they do the job of backup time 
> sources without complaint.  Tardis has got the monopoly on these clocks 
> at the moment along with Galleon but Tardis really doesn’t match up to 
> NTP other than the fancy user interface.  NTP pointing at Tardis as a 
> synchronisation server isn’t the world’s best combination either.  
> Tardis seems to claim it is perfect somehow and pretty much stops NTP 
> from selecting any other sources, even our very good GPS ref clocks.  
> This driver will stop all that and improve my time network without 
> spending serious money. 

Is Tardis running as a SNTP Stratum 1 server?

> At first glance, all is appearing well.  The clock is sitting on my desk 
> ticking away quite happily and next week is going onto a test server in 
> the server room for further soak testing.   It has also halved the 
> typical offset I tend to see between the ARCRON clocks and a GPS 
> reference.  Still not quite as good as that quoted by other people for 
> Linux/Unix so it would appear that Windows is still having the last say 
> in the matter.  But still well within the precision of this particular 
> clock.

Maybe you still have some bugs in your changes or your setup is incorrect.

> I have tried to keep as much of the non-Windows bits as intact as 
> possible – I am in no position to judge if they are good or bad, only if 
> they work with Windows or not.  Where necessary I have simply hidden or 
> swapped incompatible code inside some #if macros.  I am busy tidying up 
> the code at the moment but have left most of the previous author’s 
> synchronisation comments and code intact.

That's required since it needs to run on multiple platforms.

> I have had to work from 4.2.5p113 because 4.2.4p4 doesn’t compile in my 
> environment.  I am assuming you would be happy with that?

Since this is an enhancement it *should* go into -dev

> I do however have a couple of additional changes that I need to make and 
> I am not totally sure about how to keep them reasonably compatible with 
> other OSes.  Would you be able to help me and provide advice, or at 
> least point me in the right direction?

It depends on the nature of the changes. I cannot answer that without 
seeing the code.

Which driver are you modifying? Is this the type 27 MSF_ARCRON?

Danny


More information about the hackers mailing list