A set of slides from the military office responsible for GPS describes
this problem as well as Y2K compliance issues at:
http://www.laafb.af.mil/SMC/CZ/homepage/y2000/y2k/tsld001.htm
According to slide 4,
GPS system time counts weeks from midnight 5-6 Jan 1980 in modulo
1024 (0-1023)
No year or century data is included in the Navigation Message -
Subframe 1 (Word 3)
GPS week rolls over from week 1023 to week 0000 at midnight 21-22
August 1999 and approximately every 20 years thereafter
Clearly documented in ICD-GPS-200, para 3.3.4(b)
GPS receiver manufacturers responsible for their designs conforming
to ICD-GPS-200
So the problem occurs the night of August 21, 1999, when the week counter
rolls over from 1023 to 0. This is part of the GPS spec but some
receivers may not have been thoroughly tested for proper operation during
this transition.
The other slides indicate that the GPS satellites themselves have been
verified to have Y2K and rollover compliance, but some of the ground
support software is on older mainframes and has some Y2K issues.
I gather that the ground software would not affect the operation of
the satellites in terms of giving out good data, but would affect the
military's own use of the satellites.
Hal