[LEAPSECS] Problems with GPS?
magnus at rubidium.dyndns.org
Wed Feb 10 18:10:19 EST 2010
Matsakis, Demetrios wrote:
> I have no privileged information and GPS receiver programming is not what I do, but I also have
> never known the ICD200 to be incorrect. Perhaps ten years ago, two timing experts separately
> informed me of great frustration they had writing their own GPS receiver software, but it turned
> out they were following an incorrect unofficial summary of the ICD. If anyone is aware of
> ambiguous wording in the ICD, as claimed below, they should notify those groups also listed at the
> very bottom of this email.
ICD200 used to be a little hard to come by, but these days it's very
easy to find and follow. There are a few details that requires care and
a number of implementers practices that is not covered.
> In contrast, this listserve has passed along many stories about dumb
> programming involving leap seconds, including both GPS receiver and
> NTP software. Hopefully the manufacturers fix those too, but I'm
> beginning to think Pasteur was not entirely correct. That is, for
> some kinds of bugs spontaneous generation appears to be a valid
There are many types of bugs that may creep up. Just looking at ICD200
is not enough. One way of avoiding them is by the awareness medicine,
making people aware of a type of problem makes them look for it, avoid
it and test it again. Another way of avoiding problems is to provide
test-suits for software that can be used by many implementers. Further
testing can be done by black-box testing with GPS simulators.
Regardless, bugs will creep in and some will always be unique to the
particular environment they live in.
One aspect which can make old designs fail to work is that both ICD200
and the use of it slowly change. The use of PRN31 and PRN32 is among
those, as is the wrapping of week numbers and its handling. In the end,
black-box testing needs to be done to verify the lifetime of a
particular firmware. Firmware life-time warranties is not given and the
users needs to protect themselves and plan for upgrades before they are
bit. If vendors make long-term commitments then that is fine, but
effectively the commitments have varied greatly and customers have
suffered severely at times (including over a month of downtime of their
Firmware bugs and pure hardware problems are more probable for most
users than jamming, but neither is being cared too much about in real life.
In the end, you have to follow the first rule of buying equipment:
What's wrong with the lowest bid? Find out and discard it, look at the
next one and keep doing it until you have found one without any nasty
problems attached that you can't live with.
The problem with GPS receivers have been that overall, they just quietly
sit there in the racks and keep working and do not require attention.
They do not show up on risk analysis since they just work well. If they
where more problematic they might be looked on as a risk just as with
any other one.
More information about the LEAPSECS