Re: [gpsd-dev] Updated docs on NTP segment management

[Eric S. Raymond]

Miroslav Lichvar <address@hidden>:
> On Thu, Feb 26, 2015 at 06:01:57AM -0500, Eric S. Raymond wrote:
> > Miroslav Lichvar <address@hidden>:
> > > I think it would but how would it signal that the fields are set to
> > > not use some random garbage the client had on stack? Include a
> > > checksum?
> > 
> > That's easy.  Assuming we have the ntpd project's cooperation, we can 
> > define a mode 2 that uses a leading segment of the dummy fields to pass 
> > additional info.  Obvious tricks with unions apply.
> 
> How will gpsd know it's talking to an older ntpd that doesn't support
> the new mode and is ignoring all samples? It could check if the valid field
> is getting cleared, but that will happen also when ntpd isn't started
> yet.
> 
> I think it's important to keep it compatible both ways.

Then we'll need to checksum.  ntpd doesn't ID its protocol version, and
I distrust confiuration switches.

This is making a good argument for all-JSON, all the time...
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>