On Sun 04 Mar 2007, Kake L Pugh kake@earth.li wrote:
I was rather hoping we could get a bit of a discussion going on the list over what people are interested in, whether anyone's halfway through some code and would like some input to get them over a hurdle, whether someone's longing for a feature that someone else has had some useful thoughts about, etc.
One thing I'm thinking about at the moment is our handling of the X/Y/lat/log data; this is fuelled by my playing with GPS. (This is a bit UK-specific, I think, sorry.)
First of all, I'm a little reluctant sometimes to use OpenGuides as the canonical repository for my hand-collected geodata, since Geography::NationalGrid::GB, which we use for internal conversion, is in disagreement with the conversions produced by the conversion tool on the Ordnance Survey site[0]. (Looking at the PDF guide I downloaded from there, I _think_ the problem is that while Geo::NatGrid::GB does part of the lat/log->X/Y conversion, it doesn't bundle the datafiles that are necessary to add the final corrections - is that right?)
[0] http://gps.ordnancesurvey.co.uk/etrs89geo_natgrid.asp
So - if a Guide accepts data as X/Y coords, then when I put in the OSGB data generated by my GPS unit, the Guide will inaccurately convert those coords to lat/long, and spit out inaccurate lat/long when someone asks for the data. If it accepts data as lat/long coords, then it will spit out inaccurate X/Y when someone asks for the data.
I'm not sure what the answer to this is. There was a discussion on the london.pm mailing list recently, in which people spoke lots about releasing code, but nobody actually did.
I'm vaguely wondering if the OS would be happy with a module that let people query their webservice automatically, and if so, if it would be easy to make the internal coord conversion of OpenGuides more "pluggable", so people could choose to use the webservice module instead, if they cared more about accuracy than about being able to keep going in the face of the OS website not responding.
Another possibility would be to let people have the option of entering _both_ lat/long and X/Y, and treating them as effectively independent sets of data. If someone chose not to enter both, the internal conversion could still take place, but the data would be flagged (in the database and the output) as auto-converted.
What do people think?
Kake