Re: [nm-wg] standard timestamp format

[stanislav shalunov <shalunov@internet2.edu>]

Brian Tierney <bltierney@lbl.gov> writes:

> http://www-didc.lbl.gov/GGF-PERF/GMA-WG/papers/GWD-GP-10-1.pdf

I was looking at this document and have the following comments:

Section 4.1.1:

NTP version 4 does not pad fractional part with zeroes.  NTP uses
seconds unsigned seconds since 1900, not since 1970.  (Unix uses
signed seconds since 1970.)

Section 4.1.3:

It might (or might not) be interesting to look at the OWAMP
specification (work in progress), draft-ietf-ippm-owdp-08.txt, and how
it handles timestamps and their accuracy (in section 6.1, starting
with ``The format of the timestamp...'').

Sections 4.2.2 and 4.2.3:

The timestamp with precision and/or accuracy spec added is no longer a
valid timestamp per ISO 8601.  It would be if ``Z'' were followed by a
space.  Human readability wouldn't suffer (in fact it might be
improved).

> http://www.cs.tcd.ie/coghlan/meta/datagrid/PDPTAv7_2.pdf

The POSIX model of handling leap seconds is, of course, quite broken.
The recommendation of this paper, if I understand it correctly (to use
TAI for binary representations), isn't reflected in the first one, is
it?

This is the approach taken by the right package for Unix timekeeping.
The name of the package is, indeed, homonymous, but this setup
contradicts the POSIX standard.  It would also not be what NTP does.

By the way, the second paper says:

# [...] TAI represents time as a simple count of seconds since the epoch
# (UTC does not), [...]

Interpreted in the right way, it might be literally correct, but it
can also be somewhat misleading: both UTC and TAI representations
break down the time into year, month, etc.  Both contain enough
information (together with context) to compute the number of seconds
since a given epoch.  The difference is that for TAI, this context is
a static piece of knowledge that can be embedded in a piece of code
that will never change, while for UTC the context incorporates a table
of leap seconds between 1972 and the date to be converted into the
number of seconds.  Does this agree with the intended meaning?

-- 
Stanislav Shalunov		http://www.internet2.edu/~shalunov/

This message is designed to be viewed with 0.06479891g of NaCl.