[Geowanking] help invalidate a silly MS geographical patent application

[Andy Armstrong]

On 7 Feb 2005, at 03:40, Justin Mason wrote:
> Note, that's "base-N", not a specific base that isn't 10 ;)
>
> If you have implemented a system that encoded a lat/long pair in an
> integer pair, before the date of July 31, 2003, it might be worthwhile
> sending some prior art over to the USPTO and MS' attorneys as described
> at:
>
>   http://taint.org/2004/11/23/042513a.html

I'm sure I read a proposal for representing latitude and longitude by

* converting each to a 32 bit integer where -2^31 represents -180 
degrees and +2^31-1
   represents +180 degrees (so essentially a fixed point fraction part)
* then converting each to a string of characters in base 26 (or so) and 
using letters
   as digits starting with most significant digits
* then interleaving the resulting codes so you get the most significant 
'digit' of
   latitiude followed by the msd of longitude, then the next digit of 
latitude and so on
* then truncating the resulting string according to the desired 
resolution

This has the advantage that you can specify more precision by adding 
more characters - in fact it doesn't have to be limited to a 32 bit 
intermediate representation. Two characters are enough to express 
/some/ useful information. Also the lexical ordering of the resulting 
strings implies geographical clustering.

I'm pretty sure this wasn't an MS proposal but I can't now find it with 
Google. Anyone remember what it is? The MS proposal (I think) misses 
out on the whole variable precision benefit of this approach.

-- 
Andy Armstrong, hexten.net