[ntp:questions] Petition to FCC for accurate timestamps and NTP

Tony Rutkowski trutkowski at verisign.com
Wed Jun 6 17:20:26 UTC 2007

Hi Dave,

>The DoJ and FBI strategy apparently is to request the FCC to promote the
>DoJ position as a rulemaking issue. That is an interesting strategy, as
>formerly the carriers operated with a voluntary industry standard. The
>Feds didn't get what they wanted in the industry standard, so  now they
>want the FCC to be the hammer.

That's a good summary, although it's worth noting that some of the
history - namely that the carrier industry refused a 100 millisecond
timestamp requirement back in 1999, and it went to the FCC then
and they imposed 200 milliseconds as a compromise and placed the
requirement in the Code of Federal Regulations directly.  They also
last year reapplied the same requirements in the CFR to IP services,
so it is possible the FCC may simply say this - it's 200 ms, it's in the
CFR, providers must comply irrespective of what's in or not in any
industry standard.

Most other countries have similar requirements, only the work is
done through ETSI standards and the national authority imposes
the standards.  Those standards are now 1 second, but different
countries impose more tight specifications.

>The issue on timestamping says nothing about NTP, just the accuarcy
>requirement of 200 ms and delivery of the intercept within 8 s. It says
>nothing about UTC or NIST/USNO traceability.

You've put your finger on the key omission.  The FCC requirement
implies traceability - just as they do for frequency standards
requirements - but doesn't explicitly state it.  Some of the U.S.
standards groups have taken that "hole," and actually recast the
implied traceability requirement to be a local network clock with
no stated accuracy or traceability.  Thus the timestamp becomes
essentially worthless as the local clock could be reset to 1970
and still meet the standard.

There is a need for counsel by folks who know something about
time specifications.  Whatever one's views about the context of
the requirement, getting the right specifications articulated and
good NTP based time practices driven into the infrastructure seems
like a "good thing."


More information about the questions mailing list