Archive for April, 2007

U.S. Route 90 .. Katrina destroyed that bridge, but it’s still in Google Maps!

At Web 2.0 Expo, had the pleasure of meeting and hanging out with disaster geeks Jesse Robbins and Allegra Searle-LeBel. Jesse gave a great rapid talk on his experiences organizing relief for Hurricane Katrina and the applications to the web and technology, Failures, Disasters, & Resilient Design.


One point in his talk that really jumped out for me was about a bridge. US Route 90 travels over a bridge from Waveland, Mississippi to Pass Christian, Mississippi. Jesse established relief operations in Waveland. The entire area was absolutely devestated by Katrina, and that bridge was destroyed in the storm. However the Red Cross was directing people to travel over that bridge. When Jesse tried to let them know they were sending people to a bridge which no longer existed, they said “but it’s here in Google Maps!”. Even when he succeeded in convincing one person at the Red Cross that the bridge was gone, there was an entire organization, minus one, that wasn’t getting the message.

This is the most basic type of situational awareness that distributed, cooperative mapping can bring to disaster situations. Simply letting everyone know that a bridge doesn’t exist is not without technical challenges in a disaster zone, but we’re also talking about an organizational and social problem. The Global Connection Project which processed NOAA imagery post-Katrina, for Google Earth, is growing into something more substantial to address this.


Looking at Google Maps in Waveland today, the bridge is still there. Yes there was a recent flare up about the satellite imagery reverting to pre-Katrina .. but here we’re talking about the road data. This is vastly simpler to update, all it takes is someone at a computer about 1 minute to remove that bridge.


It’s not just Google, but every major web mapping provider that’s out of date. Here’s Yahoo giving directions over the bridge. The issue is with the data providers, Navteq and TeleAltas, whose business processes insert huge delays between reality and its representation catching up. Yes, there are efforts right now to rebuild the physical bridge, but that doesn’t excuse a huge obvious mistake from persisting for over 1.5 years.

This model of collecting and distributing mapping data is fundamentally broken. Basic geo-information about this world is too important and changing too fast to be in the hands of closed off corporations. Of course, there is another way, where the loop between users of data and contributors of data is closed (in an open way), and where the time data was collected and updated is transparent.

Comments (6)

Web 2.0 Expo, Location Intelligence

Wow, that was an intense week. Yuri’s Night, BioMapping SF, Web 2.0 Expo, Location Intelligence, NASA, AAG, and one incredible meal. Loads of old friends and new, and many motivated people.

Andrew is right, being between the conferences was an interesting place to be. Seemed like they were talking to each other and didn’t know (though Web 2.0 was a kind of massive heard rumble, like migrating water buffalo, LI like a quiet isolated group of Mountain Gorillas). We’re quite happy to be a bridge between web2.0 and hardcore geo.

The guys from weogeo were also hopping between Web2.0 and LI and rightly so .. their stuff looks awesome. Sean Gorman from Fortius One was at least looking across town and has a good account of the reception at LI to the OpenStreetMap anarchists .. though Sean Phelan of Multimap gave some excellent boosterism for OSM.

The presentations we gave covered a lot of the same ground, and are a good quick primer on web mapping. Andrew has links to the presentations.

Comments (2)

Spider vs. Bumble Bee

Spider vs Bumble Bee

Any ideas on species? Photo taken in Brighton, UK..



This next week looks to own me. Flying out tomorrow. Yuri’s Night Friday night, looks amazing. The presenting at Web 2.0 Expo and Location Intelligence. Maybe see you somewhere?


Ordnance Survey Releases Administrative Geography under Creative Commons

Marc from Geonames announces that the Ordnance Survey has released ontologies and a semantic data set of administrative classifications. Good step. The OS is a big organization with lots of history, and slowly I think it’s adapting and figuring how to operate in a new landscape.

I’ll have more to say about this, with Ian Holt at Where 2.0, where we’ll talk about how the Ordnance Survey is using GeoRSS.


Trac on Dreamhost. A nightmare but maybe

Like many developers out there, I love trac. Because it’s not easy to setup or maintain, this love is a curse. Fortunately we’re in a bright spot at the moment, here’s how I got out of the rut…

Recently my trac installs for Mapufacture and worldKit have just not been working. At first, this looked to be due to a recent configuration change for FastCGI on Dreamhost, and I requested the fix but no go. Trac then started complaining about…

OperationalError: database is locked
args = ('database is locked',)

Trac usually uses sqlite, a single file SQL compliant db. If I tried to access it on the command line, I get a similar complaint about it being locked. Killed off the fcgi processes, waited .. nothing would help except for moving the db file off the file system and accessing it from /tmp, where it mysteriously unlocked. sqlite didn’t like something about Dreamhost’s nfs, and DH support, though helpful, couldn’t resolve it.

Reading up on trac, I found out there was beta support for MySQL db backend in 0.10.4. This sounded a lot more stable than sqlite actually. So I followed the instructions here to fresh install and create a trac project, with 0.10.4dev rather than 10.3. The instructions for configuring it with the mysql backend are here.

This left me with a working Trac, but none of the old tickets and pages. So I needed to get the content from the sqlite db into mysql. Now my prior version of Trac was 0.9, pre-mysql support, and some of the column names were valid under sqlite but not mysql .. like “change” and “sql”. The idea for switching to mysql would be to “upgrade” the db with trac-admin, then dump and upload sqlite into mysql. But since the sqlite db was locked when on the nfs, I couldn’t do that.

So first, from /tmp, I dumped the sqlite db into a file. I then manually applied the db migrations. Those could be found in lib/python2.3/site-packages/trac/upgrades/. Looking in the “system” table, the “database_version” number was set to “16”, so I manually applied 17-19, and then set the db version to “19” manually.

Also had to tweak the sqlite SQL. Table names for INSERTs were in quotes, which mysql doesn’t like, so removed those. In the table CREATEs, many TEXT columns had UNIQUE and PRIMARY assigned to them, which mysql again doesn’t like, so I removed those statements too. Finally, I had to manually add AUTO_INCREMENT to the primary keys on the “report” and “ticket” tables .. sqlite assumes that any integer primary key is auto incremented, but mysql needs to be told to.

It might not have been necessary to go through all that, the trac wiki suggests this conversion method and that might work too. In any case, trac is now almost working for me. The one remaining issue is that
svn and trac are out of sync, so that updates aren’t recording in trac, not visible in the Timeline and causing problems when Browsing Source. For the moment, I’m occassionally running “resync” with trac-admin. Any thoughts there?

Comments (1)

BrightEarthProject launches on Google Earth

From Ogle Earth, I hear that the Darfur Crisis layers have been turned on in Google Earth, in the Global Awareness layer. Well done to Michael Graham and the Bright Earth Project and Holocaust Mesuem for seeing this through a long production process. I second Stefan’s praise of Google for supporting this effort, and for collecting the high resolution imagery underlying it all .. viewing the overlays, along with the aerial imagery of the places where these tragedies are unfolding, stops you cold. This is all happening in plain sight, viewable on our computers .. and it needs at the very least our attention.

I contributed a layer visualizing security incidents in Sudan, but since it’s a bit experimental and relies on the time navigation, it wasn’t included. I was given a spreadsheet with the incidents from the US State Department .. a very detached and unreal way to see these events. The visualization is based on Iraq War Coalition Fatalities animation, and tries to give some impact to the reports. Download the Sudan Security Incidents here.

Update: Primary reason the incident layer was not included was because the State Dept felt that the data collectors put too much emphasis on military incidents and didn’t account for a large number of civilian incidents. So please keep that in mind when viewing the layer. Anyway, looks like the launch is getting good coverage.


Skiing the Top of the World. Indexing the Earth.


A couple weekends ago we were in Flims, Switzerland. That’s our GPX track from one day, in Google Earth. I spent a lot of time gazing the profile of the Alps, sometimes while moving. It’s dizzying.

Most definitely that view, standing on the ground, at 46.88076, 9.204135, looking southwesterly, is unique on this earth. The profile of the peaks on these ahistoric warm and clear days stood out sharply against the sky. From this photo, it would be easy to extract that profile. From only that photo profile, could the precise location be discovered. Or .. are mountains another technology to determine location.


The more summery view from the same spot in Google Earth isn’t convincing that it’s possible.


I generated a profile for this spot at Hey, what’s that? and the result isn’t any more convincing that’s it’s easy to search by mountain.

Suppose there’s some very rough relation between the three .. and perhaps it’s just the suitable abstraction that would reveal a match. Can the ups and downs of mountain peaks against the sky be translated into an indexable number or pattern? Can we truly Google the Earth? Is the Earth Turing complete?


Mapping Twitter

I’ve been buzzing over the possibilities of mapping Twitter for a few weeks. What Twitter has done is opened a platform for lightweight mobile social messaging, that can possibly be overloaded and extended with new services. There have been numerous projects out there for location based social networking (dodgeball, plazes, many others) but nothing has caught on, for various reasons. Dopplr looks like it has niche potential here as well. So adding mapping to Twitter could be an easy hack, just as adding mapping to flickr was, and it’s happened quite quickly .. Andrew had a geo-twittering round up a couple weeks ago, though already out of date.

My thoughts have been basically, get geo information directly in the Tweets, and tailor location based services to Friends and Followers. TwitterVision has gotten huge attention and deservedly .. it’s like World as a Blog on crack. The excellent thing done here is a way to specify location directly in the text of SMS, as described in the TwitterVision FAQ. This kind of ad-hoc semantics is what machine tags did for mapping flickr.

So the other day I sent out an email to a few folks laying out my ideas. And following that, I decided to try and contact the author TwitterVision, David Troy. Amazingly, he had just arrived in London for a few days work, in offices just half a kilometer away where I’d be in London meeting on Thursday. Incredibly auspicious timing!! And meta-reflective of exactly what the technology is supposed to do. So I arranged to hook up for drinks, hung out after my day meeting observing the flows and bizarre architecture inducing behavior of the ipod styled morelondon complex and thinking about how to map such a space in OpenStreetMap. Got invited up to the check out the jaw dropping view of the offices of truphone and heard about the cheese they make on the side.

Had a good conversation and beers with David and his coding partner, interesting insights into what he’s built and how it is to build services for a twitter with growing pains. He has more in development, and twittervision seems primed to scoop this niche. I’m pondering ways to leverage all this, and Mapufacture updates as tweets seems a distinct possibility.

There’s been some blog speculation on Twitter in disaster response. SMS is one of the very least communication infrastructures that can be relied on in a disaster zone .. certainly it would be available before internet access and the power that requires. Twitter turns SMS into a social networking platform, and I could definitely see some version of that being utilised, especially with location folded in. At Where 2.0, there’s going to be a “Mass Market” demo of technology and standards, possibly a disaster response scenario, and some form of mapping twitter would fit right in.


A War of Geography

The current stand off between Iran and the UK over navy personnel boils down to extreme measures justified by bad geography. There is confusion over what the border is and where everyone was. And no, it’s by no means a defense of what’s happened, or even a criticism of the UK govt (who must be aware that their military exercises are pretty dang close to Iran anyway, and want it that way) .. I just find the ambiguous nature of imaginary lines fascinating.

There are scores of territorial disputes across the globe, yet most every common map illustrates borders as thick heavy lines. From porous internal EU borders (driving from Germany to Austria is like crossing a US state border), to heavily military-level fortified peacetime US-Mexico border, to the wide borderlands between Haiti and Dominican Republic .. what visual representations can really be faithful to the true situation on the ground?

In my ideal world, the ships of the two opposing sides dock next to each other, break out laptops, share wifi if necessary (“our WEP key is DEATHTOBRITAIN”), and OpenStreetMap until they were satisfied .. turn a naval war into an edit war .. not that OpenStreetMap isn’t occasionally just as conflicted as the Middle East.


The UK released the last GPS coordinates of the captured ship as “29 degrees 50.36 minutes north, 048 degrees 43.08 minutes east”, or 29.843333, 48.718889. The BBC has gone into more depth on the disputed positions and geographies. Yes it’s highly dubious that the Iranians changed their reported GPS coordinates to defend their position .. we’re not exactly in an urban canyon here with a bad signal.


On gmaps, you can see a bit of what the core of the dispute could be about. The delta of the Shatt al-Arab is quite naturally full of sediment, which does shift over time. The closest that Iraq and Iran have to a treaty over their precise borders in this area is a 1975 agreement that the boundary along the Shatt al-Arab extends along the median line of land exposed in extreme low tides, which the mud banks in what is commonly labeled open waters. The UK has called these some areas maritime waters in their communications, though technically they wouldn’t be under this definition. The thing about deltas and mud banks is that they shift, so they 1975 agreement includes provisions to resurvey this mud every 10 years, which actually hasn’t happened.

Craig Murray, a controversial former UK embassador, has been arguing that the situation is extremely unclear because Iraq and Iran have never formally agreed on the border in the delta. Craig Murray has negotiated territorial disputes before, and in two fascinating posts and comments discussions (if you ignore the annoyingly political Blair is a liar type stuff), he gets quite geeky on the geography and diplomacy. Territorial waters are generally agreed to extend 12 nautical miles from permanently dry land, unless that overlaps with the same from another country, then they must come to some agreement. The border in this case generally should follow some median line, and insure that both countries have navigable waters in the area. The coastlines in the area are quite convoluted, and navigation is tricky with all the mud banks. To come up with that median line, both parties must agree on a set of reference points to define the median line, and some of those points could be land only above sea level periodically, so those are assigned fractional weights in the calculation.

The International Boundaries Research Unit has published an excellent summary of the Iraq Iran Maritime Boundary, with clear detail on the geographic and political questions. AP also released a summary of the border murkiness. In summary, the boundary would be pretty clear if we lived in a rational world, though we clearly do not.

This isn’t the first time the Iran-Iraq border has led to conflict. Heck, when have borders not led to conflict .. absolutely harks back to maps as a military technology. Anyway, in 2004, a similar incident occurred directly on the Shatt al-Arab. In part, the dispute sparked the Iraq-Iran War in the 80s. And it’s been going on since a peace treaty signed in 1639 between the Persian and the Ottoman Empires. Wikipedia has a good account.

Comments (2)