Monday, 26 March 2007

Reality Check

We're now a week and a half into this exercise and the first reality check has occurred - a colleague has been off sick all week, meaning that I've had to cover his load, meaning that the fun parts of this job (like SDI indulgences) get less attention. I guess this won't be the last such distraction but it is annoying to get enthusiastic responses to the initial invitation and not have the time to respond well.

At this stage I have about 30 expressions of interest, from as far afield as South Africa, Zimbabwe, the US and UK, from within the UN as well as academe, industry (even a cement company!) and the conservation community. Not a bad start, but I'm hoping for more a we make progress and word spreads.

My colleague John Mugwe and I last week even struck our first small blow for Real Interoperability when we visited our colleagues at FAO/ SWALIM (Somalia Water and Land Information Management System) to install the geoserver web feature service. John argues that we delivered 100% of what we promised, with SWALIM staff now able to install and run geoserver against the shapefile library (also available for doenload from their GeoNetwork node). Me, I wanted to get them up to a spatial database and pushed for installing PostgreSQL and PostGIS which all went fine, right up to the point where we actually tried ingesting their shapefiles using qGIS. Whack! Instant embaressment. Oh how we tried and struggled and seated and cursed, all to no avail. Shamefacedly did we crawl away from the scene of defeat. Humbly and contritely did we check and recheck and analyze and re-analyze what could possible have gone wrong with something that between us we've done dozens of times.

Back at the office we ripped out, reinstalled, tested and retested and just could not re-create the problem. Until, that is, we asked SWALIM to send over one of the shapefiles we'd failed with. And, Lo!, the problem resurfaces. Obviously those FAO losers have bad data. Lets test it with and import into ESRI's own products to see at what point the process stops. Except it doesn't stop. Re-export the data and try again importing to postrgreql using qGIS and it flops. Until, that is, the import-and-re-export cycle is tried again but this time not using the original shapefile name. And thus did we successfully import to postgreSQL. Why, oh why, were we struck with such problems. As it turns out the culprit was not bad software, or bad data (the FAO teams are not such losers after all), but bad file names. It seems that any name with a hyphen in it is poison to postgreSQL when it's setting up the tables. Sigh.

Now all we have to do is work out how to unsully our reputation with FAO. But at least tomorrow, perhaps, we'll be able to show the first interagency interoperation here in Nairobi.

No comments: