Dunno - would be phone specific.  Would probably be easier with a
Java-based or Windows Mobile phone.  You could probably set something
up that worked when the power came on rather than having to rig up a
serial input.  Sod all use if they trailer it away though.  Not my
area, but you might be able to read the cell information depending on
your network to get an idea of whether the phone is at home or not.

Anything else requires GPS, which adds cost, serial comms and needs
view of the sky.

Prolly a better idea in some senses to do so IMO.

The VMware session is fired on boot - I'm happy that I can just kill
Windoze like any other app :-)

Nothing yet - just running it as a car PC - for satnav and music etc.
I figured I really need to 'just do something you lazy sod' rather
than waiting until I had everything I need.  I can add any interface
to it as I need to.  The basic code is fine, but the IO has jet to be
decided - and what plugs into the IO, obviously :-)

the motorway posts have numbers on 'em too - you could pick a suitable bit
of motorway driving, hit the trip meter as you pass a post with a known
number and then drive e.g. 100km or something (or more, if you're going
further on the same motorway) and then note the reading.  In fact, for
better accuracy, you could pull onto the hard shoulder, park beside a post,
zero the trip, then drive, then pull onto the hard shoulder by a post near
your exit point, and note the mileage on the truck milometer as compared to
the distance on the m-way markers.

if you can do about 200Km on the same road, the measured distance should be
accurate to well under 0.1% - the markers are 100m apart, so over 200Km the
distance between markers is 0.5% and even if they markers are not measured
with pinpoint accuracy, on that sort of a scale you can get well under one
tenth of a scale division[1] accuracy.  I'd have thought that would get you
close enough, really.  you should be able easily (especially if you repeat
the exercise on several different motorways) to get the accuracy down to
0.05% by suitable statistical analysis, maybe better than that.  You can
position the truck within a metre of the marker post, I reckon, easily
enough if you stop on the hard shoulder beside it...

[1] cue discussion about accuracy of readings: when I was doing physics in
uni, the bloke teaching us about experimental technique held that if you
have a scale with divisions n, and measure a quantity x using said scale,
your accuracy was always +/-(n).  I held than in general you could easily
get +/-(0.5n).  There are special cases where this doesn't apply, but as a
general rule, your starting point will be chosen to match a scale division
to high(ish) accuracy.  

Suppose, for example, you measure a steel bar, which has an actual length of
123.4cm using a ruler with 1cm divisions.  You choose to put the ruler so
that one end of it (or a chosen scale mark) lines up as closely as you can
get it to one end of the bar.  you then judge the length by looking at the
other end, which falls between 123 and 124 on the scale.  In this case, it
falls somewhere approximately half-way between the marks, and therefore you
can quote the measurement as 123.5 +/-0.5cm.  The lecturer bloke was adamant
that you'd quote it as 123 +/-1cm.  If the reading was obviously a lot
nearer to 123 than to 124, you'd quote 123.0+/-0.5cm.  You can argue that
you have an error at both ends of the measurement but for the vast majority
of measurements, you wouldn't do that, you'd move the scale and reduce the
error at the starting end to a minimum.  Even if you can't actually do that,
for almost all scales, it's possible to judge to the nearest half-division
by eye.