Single Point of Failure

Posted by Zverik on 29 November 2012 in English (English)

On 27th Steve Coast wrote in his blog a big speech about people who take the sole responsibility for important elements of infrastructure in crowdsourced projects. He tells of points in OpenStreetMap which are controlled by indispensable people, and other projects members acknowledge that they have no idea what to do if those people would quit. It's time, Steve says, for them to leave, so we could «find out how we distribute the workload now while we only have 900,000 registered accounts instead of 9 million».

This is a familiar subject for reliability experts, the Single Point of Failure talk. In crowdsourced projects such points are people, of course. And indeed, several years ago only one man worked on Potlatch, another one made OWL, the mapnik style was controlled by third, and this chain of indispensable people went on and on. But Richard Fairhurst correctly notices in comments that today in all those control points communities have grown, and indispensable people are not indispensable anymore. Actually, I can name only one SPF in our project from the top of my head. And it's Steve himself.

Who is contacted by media when OSM has another victory over some proprietary cartographic service? Who is deemed visionary, who opens every State of the Map conference? Who our project is associated with by outsiders? Is it someone from the press contacts? No. It is the Chairman Emeritus of the OSMF Board and the Inventor of OpenStreetMap. What is his recent archievement in the project? On his site there is a list: "founded OSM" (2004), "started SotM" (2007), "co-founded Cloudmade" (2007 again), "kicked off mapstraction" (2010, take a look at djeo btw). Which are his latest two projects? FrontDoor, which allows to fix houses' entry points using Bing imagery. Last month James Fee asked Steve, where are the results. He evaded the subject. And OpenGeoCoder, for which Steve specifically chose not to use OSM because of Share-Alike license. His last edit to the map was in February 2011.

You know how most active people in our community respond to imports. Only recently there was a major flame war in mailing lists concerning French cadastre buildings. Few consider TIGER experiment in USA to be successful: last week's Operation Cowboy's main objective was fixing data imported five years ago. And what major problem does Steve identify in his yesterday note, what solution he proposes? Addressing, of course. US has too many buildings, and it is impossible for the mapping community to collect all the addresses. We just don't have enough time. And because there are vast emptiness in OpenStreetMap instead of addressing information, it is unsuitable for geocoding. And the situation is even worse in Europe and Japan. I guess no one ever succeeded in finding an address there. What could we do to save the project? Let's take TIGER data (which is, of course, horrific, but there is no other way), slap it into OSM and tell the community to fix it! Better than nothing, anyway.

Now I should remind you that Steve Coast still represents our project. Not Germans, Jochen and Frederic, whose articles I adore. Not Englishmen, Richard, Harry and Matt, messengers of reason. Not Simon Poole, current chairman of the OSMF Board, and not Richard Weait, an activist of Communications Working Group. Steve Coast. Only in October he was interviewed by James Fee, editor of the popular GIS-related Spatially Adjusted blog. In September he opened the State of the Map US conference, in November he told of his OpenStreetMap involvement to Czech GIS-portal. To this day Steve is a face of our project, despite being irrelevant. It was fun a year ago, but today I think he is starting to be harmful to project, slowing it down.

"We have points of control which are fully owned by single individuals. The people around them readily acknowledge that they don’t have a clue what we would do if they quit. I say that’s terrible. I say they should quit and we should find out what we have to do." Steve, who did you write this about?

(this is a translation of SHTOSM entry, and I'm sorry if something is translated inadequately)

Comment from Sanderd17 on 29 November 2012 at 11:41

Well, all code we use on the main site is open source. So even if some project is only controlled by one person, he can't withdraw his work.

Of course, it would be a huge task to replace such an "indespensible person", but it doesn't need to happen immediately.

But you are right that there aren't many indispensable people left. I also wouldn't count Steve as indispensible. As far as I know, he has little influence on media (at least in German articles, it's always one of the Germans being interviewed), and the SOTM talks are easy to replace.

I also don't agree we're not doing well with the addresses. Yesterday, I needed to search an address of a friend, so we could work together on a project. I just needed to give in the address in OsmAnd, and I was immediately send to the right house.

And no, I haven't added that address myself. I have never been there.

A year ago, this wouldn't be possible, but we're moving fast.

I'd say, if you don't like what Steve Coast is saying, just ignore him. It's not like he's that important.

Comment from Chaos99 on 29 November 2012 at 12:54

I've already written an quite moody comment on his recent blog post about importing more TIGER data (

But your post shows: this was probably wasted time. I don't know what his motives are right now, but helping OSM move forward doesn't seem to be part of it.

He probably just hasn't found an address he was looking for.

Comment from Rovastar on 29 November 2012 at 13:14

From what I understand there are many dev but only 1 admin.

Comment from osmisto on 29 November 2012 at 15:57

Chaos99, nice comment.

But please, don't forget that Google, Bing, all worldwide mainstream map providers look at USA as on their primary target. They spent much more resources to improve USA map (3d view, streets view, birth view, whatever) then for other world. New candies for USA firstly, all other world at last.

So, may be, OSM in USA is so poor because competitors are so amasing. They free, they sweet and they satisfy their users enough to make them happy and stop them from looking for alternatives. Tiger import may be is not the biggest cause.

Comment from Andy Allan on 29 November 2012 at 16:23

Rovastar - that's not true, there's are a group of sysadmins, all with the same access levels.

There's also difference between "single point of failure" and "single point of authority". We try hard not to have any single points of failure (e.g. someone who knows a password where nobody else does) but there will always be single points of authority - e.g. someone who is the maintainer of a piece of software, or the chair of a working group, or whatever. Not a dictator, just someone who makes the final decision if consensus can't be found any other way. And of course, in a volunteer organisation there are some tasks that only one person is willing to do - a "single point of enthusiasm", perhaps.

All of these things pose different challenges. Bundling them all together makes for better blog posts, but doesn't help solve any genuine problems.

Login to leave a comment