RELEASE_NOTES for PTPd

v1.0.0, 9 August 2008:

The long over due 1.0 release. A few small fixes, the most notable of which is a
fix to check the revision of an IEEE-1588 message to prevent PTPd from
interfering with IEEE-1588 v2 devices on the same network. There are some extras
thanks to contributors, including an updated MPC8313E patch, a QNX patch, and a
man page. PTPd v1 will be branched in the repository, and work toward IEEE-1588
v2 will begin on the trunk.

v1rc1, 17 June 2007:

This release fixes a regression in v1b5 that broke unicast support. It also adds
a PTPd patch posted on the website for 1588 hardware support on MPC8313E
processors.

v1b5, 3 December 2006:

PTPd is now ported to *BSD. Also, there are small fixes for 64-bit
architectures, better interface search on Linux, and disallow master mode if the
stratum is 255 for 1588 spec compliance. There is a fairly large change to
separate the internal time representation from the external IEEE 1588 time
representation. This significantly simplifies the time arithmetic functions, and
it should allow PTPd to work properly after 2038 despite the fact that the Linux
clock will rollover. PTPd should run seamlessly through 2038 as a slave, but to
run PTPd as a master after 2038, you need to set the 'half epoch' flag with the
'-h' command line option. Finally, there is a change so that PTPd no longer
needs a kernel modification to retrieve send time stamps. PTPd now retrieves
send time stamps from looped back received messages.

   -kendall


v1b4, 28 June 2006:

More fixes, including some to make PTPd work with the PTPManager program. The
default UTC offset has also been changed again. Now it is zero to comply with
the 1588 spec, which says that unless a clock is otherwise synced to UTC, it
should report a UTC offset of zero. The '-t' option (don't adjust the clock) has
been fixed so that it also prevents zeroing the frequency at start up. The
ability to query a network with management messages has been added with the '-k'
option. Finally, PTPd's timer and polling mechanism has been changed to allow it
to respond immediately to received messages.

   -kendall


v1b3, 6 March 2006:

There has been some useful feedback since the last release. This release
contains some requested bug fixes including an update of the default UTC offset.
This release also adds a requested slave-only mode. Thanks to everyone for the
input.

   -kendall


v1b2, 8 December 2005:

There has been no feedback from the previous release, so this release does not
contain any significant changes to the source. This release is just the few
minor source changes still sitting in CVS, along with some cleaning of the extra
bits like the documentation and tools. Also, I have removed the 'parpp' tool
from the distribution because its use is not straightforward, and I think it may
be more confusing than helpful to casual users. Finally, I included a PDF of the
paper I wrote for the 2005 Conference on IEEE 1588 to make up for my languishing
source code documentation effort.

   -kendall


v1b1, 18 October 2005:

The site seems to be getting some downloads, so I want this latest code out
there. There are only small changes. The number of foreign master records now
has a command-line option. Also, the the 'Q' randomization of Delay Request
messages is essentially disabled because I realized it was not implemented
correctly. This is no big deal, because PTPd's Delay Requests do not occur in a
consistent part of the Sync interval window, so they are implicitly randomized.
Anyway, PTPd has been plug-fested a couple of times, and everything seems to
work well, so I am comfortable calling this a beta release. Hopefully people
will try the program out and help shake out some bugs.

   -kendall


v1a4, 24 September 2005:

A fair amount of changes, most of which I do not recall. Fixing support for all
of the management messages is one that comes to mind. A fix that sets socket
reuse is another one. Also, there's a fix to the initial clock reset that
allows it to start up very close to master clock time. In general, this release
is more stream-lining and tightening up core functionality. I think that the
protocol stack and the clock servo are feature complete for v1. I will be
testing this release at the IEEE 1588 conference in Zurich. After that, I will
probably tag my next release as beta.

   -kendall


v1a3, 17 August 2005:

Lots of changes. PTPd functions essentially the same as before, but the code has
been reorganized and trimmed. Also, PTPd now has a real clock servo, and there
is a hook for kernel send time stamps. The stock Linux kernel does not have
a send time stamp, so you'll have to implement your own. I'd be glad to provide
instructions how or kernel patches if anyone is interested. PTPd has shown the
ability to hold coordination solidly within 10us running on a 66MHz M68k.

   -kendall


v1a2, 4 June 2005:

PTPd can now be built with 'make'. It also includes a little bit of usage
information in README. This release also fixes a broken clock servo that
caused a slave with a time greater than master clock time to run away from
master clock time. Finally, it includes an updated 'parpp' that hopefully
can be used to precisely quantify the precision of clock coordination.

   -kendall


v1a1, 25 May 2005:

This is the first public source code release of PTPd. The source has been
compiled with gcc and tested on an x86 Linux machine. The code may work on
big-endian machines as well, but it will only work on 32-bit Linux machines.

This release also includes a kernel module that generates a pulse-per-second
(PPS) on the most significant data bit of a PC parallel port. The PPS
generated by PTP systems can be used to verify their clock coordination. The
included PPS generator is not precise, so it does not provide a good way to
quantify the time coordination precision.

All of this code is a work in progress. Please feel free to mess around with
it and use it as you wish. While I can offer no support, I welcome any
questions/comments on porting, bugs, features, or anything else. One
exception, please do not ask me what PTP is. One can learn more about PTP
from the IEEE 1588 specification, and one can join the LXI Consortium to
work with the companies using PTP.

Finally, this release is not intended to be a functional PTP system
(although one can be built from it using 'gcc ptpd.c -o ptpd'). It should
not be too hard to figure out how to use it if you dig into the source.

   -kendall