I just love taking shortcuts through entire blocks, especially when walking. An ideal map shall facilitate that by clearly indicating any closed areas.

And I think it's better to overdo than to deliver a false promise: later someone else will map any holes that are lawfully used by locals.

Most private lots are diligently closed from public spaces and areas around high-rise buildings. But, won't the map be too crowded with all those fences? Indeed the fences that delimit one private lot from another -- though they might be visible in satellite photos, too -- have no practical use on the map. Only the perimeter is important.

The last changeset that implements aforementioned ideas rendered as expected and looks nice. Surely not too crowded. On the other hand, the entire block is rather gratifying in that regard. Another block that I've drawn from Bing on Saturday, consists almost entirely of private lots; only time will tell.

A bit more free time during holidays. Only 15 raw traces from year 2009 were remaining, so it's a challenge to finish them! Cleaning up, now, was a rather quick job. Mapping may last into January.

Almost half of traces from this year are already mapped.

Overall progress: 230 out of 270 done. (160 MB / 200 MB by size)

Another good news is that my crude logger application is now able to configure the SiRF III chip automatically. So further traces will definitely have Track Smoothing disabled and will catch EGNOS more often. Doing the same manually with SirfTech is tedious (and practically impossible if already driving).

Another source of GPS data is available. Outdated, too :)

I bought a different receiver, Holux GR-236, three years ago for some EGNOS support (this is a true SiRF III unit after all), and compact size. My smartphone, HTC TyTN, already was a good candidate to collect raw NMEA data via Bluetooth. Just think about those gigabytes on the memory card!

Initially SirfTech was enough as a logging application. A year ago I made my own logger, an extremely simple thing with everything hardcoded, and therefore rock stable at last.

To preview NMEA stream, GpsBabel was also OK. Back then it required you to choose a single particular sentence from which the data will be extracted. That means some less popular parameters like "sats", "hdop", etc are lost. Nevertheless I always wanted to see all NMEA content, up to the current constellation even, in a nice tabular form. Almost coincidentally I started coding a parser, basically as some OOP exercise; though the project was perpetually stalled, some minimal version began to work just recently.

So it is now possible to convert those ~60 additional logs (40 MB in tabular form) to GPX. Just in time because there are almost none from year 2007, and the main data set from that year is finished, too.

This time file names will be simply timestamp-based and not numbered.

Accuracy should be worse, though I didn't test that.

In the car, the receiver sits as far as possible under the windshield, there is no coating or active antenna on the glass, however that's still obviously worse than the roof.

While on foot, I keep the receiver in my satchel that has a nice outside compartment for small mobile phones. That way the antenna almost always faces horizon (not zenith), which is not good but still better than a complete chaos in a pocket or elsewhere. Those patch antennas are not omnidirectional, you know.

On the bike there is a perfect place: the handlebar. Held with a few rubber bands, of course.

Potlatch now has the "offline" mode that from the first time helped to survive a network problem. Great!

Another out-of-order trace, #87. My first spontaneous attempt to map a whole village (traces only so far). The village was small and passengers didn't object :)

I decided to exclude our main motorway (A1), except perhaps its links and crossings, from future traces. With the existing ones it looks pretty neat already, lots of possibilities for fine-tuning. Not a big issue as much data is removed already before uploading.

Apparently a preview of multi-segmented GPX may show some segments as joined, with inadequately long links that would confuse mappers. Luckily Potlatch doesn't have that issue, as far as I checked.

Still lots of tracks (almost 70) to prepare for OSM. However a recent weekend trip revealed a serious blank spot for which my recent memories (among with the track, of course) appear to be the most important factor. As written notes may not help after a few months :) , I uploaded #084 just after #012 and started mapping over it. #013...#083 should follow when I find enough time.

I've just finished a converter so now all those megabytes of traces in a custom format can be shared. I don't mind about the Public Domain license.

The receiver is a Haicom HI-204S, my first one. It was purchased in 2005 mainly for a few hobby electronics projects, one of them (and eventually the only one) being a logger. That one is rather of old-school fashion: 8051-based, 256KB of EEPROM for the track, 6 V 4 Ah SLA battery for both receiver and logger. Heck, the whole bunch weighs almost 1.5 kg. A packed binary format holds about 15K records. Being a purist, I keep the receiver on the car roof -- or on a PVC frame protruding from the backpack (cheers, Trimble!).

On the PC those tracks are archived in CSV format. A converter to PLT format exists long ago because OziExplorer is my workhorse. For OpenStreetMap I needed another one. Of course PLT->GPX using GPSBabel was not worth it as important bits like HDOP are already lost.

A bit earlier I coded a diff utility that updates my CSVs given differences between an original .plt and a censored .plt. So now I'm additionally able to weed out data that is obviously wrong, outdated, misleading, needlessly static, or considered too private -- and present it as GPX, even with support for separate track segments.

Hopefully there won't be any nasty errors in the resulting format. SAXCount.exe from xerces-3.0.0 doesn't complain but who knows...