Re: [gpsd-dev] [PATCH] Time Service HOWTO: Clarify state of RFC2783 on NetBSD.

[Greg Troxel]

"Gary E. Miller" <address@hidden> writes:

> Yo Greg!
>
> On Tue, 26 Aug 2014 08:46:16 -0400
> Greg Troxel <address@hidden> wrote:
>
>> Note that RFC2783 support on serial is ancient, and works with ntpd,
>> but that gpsd's PPS code does not work with it.   Note the addition of
>> USB-serial RFC2783 support in NetBSD 7.
>> 
>> Add a missing "Linux" qualifier in discussion of Linux KPPS.
>
> I disagree with that addition of Linux to KPPS.  More than Linux
> implements RFC2783.

The paragraph where I added Linux was about Linux.  Yes, RFC2783
works on other places.  But it's not "KPPS", which is a Linux-specific
term.   Arguably this should be restructured to explain in terms of
standards and then per-OS how to configure the system to provide it, and
then I think we'd both be ok with it.

> As you note NetBSD has added their own RFC2783, finally.

That's not correct.  NetBSD has had RFC2783 support (really, a late
version of the internet draft the became RFC2783) since about 1999.  The
only recent change was to hook it into usb-serial devices.  The
sys/timepps.h header has:

/*
 * This header file complies with "Pulse-Per-Second API for UNIX-like
 * Operating Systems, Version 1.0", draft-mogul-pps-api-05.txt
 */

which I should check over vs the published RFC and change the comment.
But I don't think it actually really changed form -05 to RFC.

> Sadly RFC2783 does not specify the entire API.  If NetBSD needs some
> tweaks let us fix gpsd, not document that we gave up.

I don't follow.   RFC2783 does specify an API.   Reading the gpsd code,
it seems to make extra assumptions beyond the API, which makes it not
work.  So certainly we should fix it -- my comment was meant to note that
it does not work, not that it never will.  Perhaps that should be
clearer.

pgpvlqJvSkiLB.pgp
Description: PGP signature