[time-nuts] sub-minute time-precision in court-case

Jim Lux jimlux at earthlink.net
Tue Sep 3 10:56:47 EDT 2013


On 9/3/13 7:07 AM, Poul-Henning Kamp wrote:
>
> It is very rare to see courts deal with time precisions less
> than minutes, but it seems to have happened in this case:


Interesting (and of course, this case has been in the news recently)..

In this case, all the messages were presumably handled by the same 
carrier, so the issue of skew in timestamps is negligible; they're all 
presumably running off the same clock.  I wonder, though, whether this 
is always the case. Yes, the cell companies have accurate timing at the 
cell site level to do a variety of things, but I could see that getting 
lost along the way to the "meta data logging" process.  The time stamp 
might be "when the message arrived at the logging process".

And, different companies might have different standards for how those 
timestamps get applied and their accuracy requirements.  I can see how a 
company might have a legacy billing system that, say, works in 0.1 
minute chunks, so a 6 second random scatter is inherent in the system.


Given that cell companies know the location of the radio with at least a 
few tens of meters, it would be interesting to see statistics of text 
messages sent/received while on freeways.  Indeed, one does not know 
that the person with the phone in question is a driver or passenger, but 
simple observation (at least in Los Angeles) shows that the vast 
majority of cars are "single passenger"  (alas, even in carpool lanes, 
there's a significant number of them).

I do know they calculate this sort of thing, because they use it to 
determine where to put new cell sites or change capacity.  The question 
is whether the information is available in any sort of useful form.  It 
could be some horribly ad hoc process where an engineer gets a bunch of 
text files, and they load it into Excel spreadsheets to process it. 
It's not like they necessarily do site planning on a minute by minute 
basis, so a manual process that takes a few days is plausible.




>
> (from: www.judiciary.state.nj.us/opinions/a1128-12.pdf)
>
> 	Best stopped his truck, saw the severity of the injuries,
> 	and called 911. The time of the 911 call was 17:49:15, that
> 	is, fifteen seconds after 5:49 p.m.
> 	[...]
>
> 	texts [...] exchanged while Best was driving:
> 	Sent      Sender  Received Recipient
> 	------------------------------------
> 	[...]
> 	5:47:49   Best    5:47:56  Colonna
> 	5:48:14   Colonna 5:48:23  Best
> 	5:48:58   Best    5:49:07  Colonna
> 	(5:49:15 911 Call)
>
> 	This sequence indicates the precise time of the accident -
> 	within seconds of 5:48:58. Seventeen seconds elapsed from
> 	Best's sending a text to Colonna and the time of the 911
> 	call after the accident. Those seconds had to include Best's
> 	stopping his vehicle, observing the injuries to the Kuberts,
> 	and dialing 911. It appears, therefore, that Best collided
> 	with the Kuberts' motorcycle immediately after sending a
> 	text at 5:48:58.
>
> Nowhere does the opinion mention if the timestamps were taken on
> the same clock or if the two clocks were synchronized.
>
> Best was a volunteer fireman, but I still find the seventeen
> seconds slightly incredible.

Based on the behavior my cell phone (sending texts in signal denied 
areas), I assume the time stamp is actually the "time when message 
processed by cell site", which could be many seconds (minutes, hours) 
after pushing the "send" button.  I've had text messages queued in my 
phone that get sent when I land and turn my phone back on.  More than 
once my wife has gotten the "they're closing the door" text from me when 
I landed.

Well, Best did take 35 seconds to respond to the text from Colonna.  I 
think we can assume that the "incident" occurred slightly after x:58 
(although I suppose the sequence could have been

keypress
keypress
keypress
<sound of impact>
keypress
<send message>

Zipping along at 30 mi/hr (14 m/s), 10 seconds is quite a distance.

This was about a half an hour before sunset on that date.
There was a turn in the road involved, etc.

What didn't show up  in the record is the lat/lon estimate for Best's 
cell phone.  I would assume that in 2009 they were logging this as well, 
but maybe the data was not in evidence (the lawyers may not have wanted 
it.. that's the frustrating thing about reading appellate cases: you 
have to work with the data as presented at the original trial)

>
> The seventeen seconds are somewhat material to the ruling, but
> not a decisive factor.
>

I think it's more the back and forth of messages just before the 
incident that are at issue: they imply that there was a "conversation" 
of sorts going on, and that the young lady may have had knowledge that 
he was driving at the time (which ultimately is what this case is all 
about).

I'll bet the marketeers at the cell company have all sorts of models of 
texting behavior among people, just waiting for the ability to insert 
ads of the appropriate type.  I can see an analysis of "should we go out 
to dinner", "Sure, where", and the ad pops up, "why not eat at Dave's 
Bar and Grill, would you like me to make a reservation?"





More information about the time-nuts mailing list