OpenStreetMap

A Rant: The Way Beyond Craftmapping That Nobody Is Talking About

Posted by bdiscoe on 25 September 2016 in English (English).

When I read Michal Migurski’s recent post robots, crisis, and craft mappers, I was really baffled and concerned. I am a fan of Migurski; he’s a good person and a smart guy. But the content of this particular blog post was really off. I had hoped it would pass with little notice, but I can tell from the #craftmapper T-shirts at SOTM that people actually paid attention, so sadly I feel compelled now to rebut, and hopefully offer some useful perspective as well.

To get something out of the way first, I am absolutely a “armchair” or “craft” mapper, and an addicted mapper, averaging ~5 hours a day mapping for the past 3.5 years; by my own estimation, there are only two human OSM accounts (katpatuka and Heinz_V) with more node/way contribution. (Also, shoutouts to AndrewBuck, Stalker61 and ulilu!) I care passionately about the map, I’ve been in geo since the 90s, and I’ve been inside Google to see how mapping actually happens at scale.

My OSM Heat Map

To start with, he writes:

The OpenStreetMap community is at a crossroads

Arguably, no it isn’t. It is actually on a stable trajectory, with no major shifts likely.

I see three different movements within OpenStreetMap: mapping by robots, intensive crisis mapping in remote areas, and local craft mapping where technologists live

Actually, no. “Robot” mapping is a perennial project of AI zealots, not a movement, and cannot and will not produce acceptable data (for reasons way beyond the scope of this rant). At best, it is another way to produce yet more controversial imports of dubious quality. Crisis mapping is now well-established for many years, not a new or dynamic trend; same with local or remote “craft” mapping, i.e. normal OSM contributors; not a movement, and not new.

The first two represent an exciting future for OSM, while the third could doom it to irrelevance.

This is saying that normal OSM contributors, the ones that have and continue to build most of the map - and the great majority of the quality map - are “irrelevant”. This is really, 100% wrong.

Historically, OpenStreetMap activity took place in and around the home areas of OSM project members

True enough, and that is still the single largest source of quality map contributions. The other parts are imports, a small amount of commercially-sponsored input, and armchair mappers like myself, tracing aerials from the places that can’t (or can’t yet) map themselves, either for HOT or MissingMaps or beyond. Together, that IS OSM, past and present, and unless Something Dramatic happens, that is also OSM’s future.

Craft mapping remains the heart of the project, potentially due to a passive Foundation board who’ve let outdated behaviors go unexamined.

I am trying to figure out how to not feel hurt by this. “OUTDATED.” The passion that drives the entire past, present and future of OSM is “outdated?”

Left to the craft wing, OSM will slide into weekend irrelevance within 5-10 years.

That’s basically saying that OSM is irrelevant today. As an opinion, that’s a pretty harsh one.

Two Modest Proposals (1) codes of conduct and other mechanisms intended to welcome new participants from under-represented communities

This sounds fine, but it seems orthogonal to the “robot, crisis, craft” framing. It seems uncontroversial to empower and support more crisis/craft mappers from under-represented communities.

(2) the license needs to be publicly and visibly explained and defended for the benefit of large-scale and robot participants

I have sat out the license wars, partly because, as a regular non-lawyer human, I cannot fathom what all the fuss is about. That said, it also seems unrelated to crisis/craft mappers, with or without AI-robot assistance to produce data for human review, who will surely be able to proceed with or without license changes.

I could say much more about this, but much has already been hashed out of the comment thread on the original blog. For example, “automation vs. craft is a strawman argument; Both - in an integrated manner!” yes obviously.

Instead, I’d like to provide an answer the question I believe Migurski is actually asking. I believe he is saying:

  1. While better in some areas, OSM isn’t on par, for the full range of uses, with maps from Google/Apple/etc.
  2. The existing approaches aren’t on a trajectory to get us there, therefore they “doom us to irrelevance”.
  3. We need something more to get us there, but what is it (robots? codes of conduct? license changes?)

The answer to this question is obvious, but everyone seems to be waffling and dodging it. I will say it: MONEY.

MONEY

To be a top-tier global map, it takes roomfuls of full-time, paid mappers, with the kind of resources and coordination that (realistically) are only found in large corporations.

  1. Clickshops. Google has them, Apple has them, any organization that wants to take OSM to the “next level” will need them. In some developing nation (for cost), with fast computers and fast networks and thorough, regularized training for speed and consistency. (In case someone is thinking Mapbox, that’s nice, but think bigger. Think 100x.)

  2. Streetview. Every station in Google’s clickshops has the entire catalog of streetview instantly available, continuously integrated into the mapping flow. Without a streetview-like dataset, you just can’t do it. I know Mapillary (+JOSM plugin) is trying, but they are not even close - you have to capture FULL 360 (cylindrical) imagery, not just hope that hobbyists were pointing their camera where you need to to look, and you need the RESOLUTION to read street names. Not even 1% of mapillary users are capturing HD 360 imagery. You can’t do it with prosumer cameras (I’ve tried). You need an expensive rig. Stop pretending otherwise.

Some company or consortium (or, in theory, government, but I’m not holding my breath) could step forward with MONEY and take OSM to that “level III/IV” Migurski (and many others) would like to see. Barring that, everyone needs to extend love to the homebrew/crisis/craft/mapathon mappers we have, because we ARE OSM’s future.

Comment from BushmanK on 25 September 2016 at 17:14

I’m don’t want to defend the idea of “robots versus mappers”, however, there is something in direction of wider usage of more advanced technologies like automated imagery classification. And AI (or classic methods of photogrammetry) can definitely help here, at least, in the aspect of filling huge gaps in barely populated areas. It is the different issue than one Michal Migurski is trying to explain, but there is strong default prejudice (echoing in this diary entry too) against automatically generated data (even in cases when it’s not about any kind of direct import).

It is quite funny that crisis mapping was mentioned by Michal as a good way, while it is a proven fact, that any local map, improved by the whole bunch of volunteers, becomes outdated immediately after HOT operation is formally finished. Exactly because local mappers, who can keep it more or less up to date, do not exist. So, after reading a statement like that, I see no credibility in any other words of its author, since it’s not even a fallacy or demagoguery (something pretending to be true), it’s just false, unreasonable, illogical.

Personally, I wouldn’t say that street view is a lifesaver. At least because the amount of important data we can extract from it is in huge disproportion with its cost. I mean, having an all-weather rig of six high-resolution cameras mounted on a vehicle to read street names, really? Leave aside that in certain types of populated areas it doesn’t help at all, you just can’t see anything like that from a roadway. So, that’s a waste of money anyway.

There is a lot of space for new methods and new technologies. Some craft mapper attitude (like certain Luddism) can get in the way of it. But if mappers are really open to learning new, more effective technologies, it could really bring the situation to the new level. And yes, to keep data up to date we need more information. But it shouldn’t be a pile of trash to sort in search of a couple of valuable grains (it’s about maps.me contribution and “community extension”) - it’s about, for example, automated track collection, which does not depend on user’s skill.

Comment from Warin61 on 26 September 2016 at 00:14

There is a perception that anyone who maps using different method from ‘me’ is doing wrong.

Terms such as ‘mechanical edit’, ‘automated import’ etc are meant by some as derogatory. Criticizing someones edits based on their method is destructive. Criticize on the content, not the method! That is the only thing that truly matters.

Methods will develop over time … don’t limit OSM input by placing unrealistic restrictions on methods used .. look at the content.

Methods! I think my best mapping is where I can mix the sources of the mapping from independent sources. My worst mapping is where I use one source. A gpx track might get imported .. but it may have some diversions .. stopping off the road for some shops, photos etc … I do come across these from time to time, easy to correct from imagery.

The requirement for ‘keeping upto date’ … well that happens where ‘someone’ notices it and takes action. If it does not get noticed … shows that information is not used. If noticed but not corrected .. the information is not valued. Only where the information is both noticed and corrected is it of value. I have had one of my edits noticed as out of date - comment placed on the node, but no attempt to correct nor notify me .. so it was not correct untill I happened on it. Out back road in Australia get maintained by grader drivers - they are dirt. If the grader driver thinks the road would be better off ‘over there’ then he puts it ‘over there’! Thus the map can be out of date very easily. Of course the road will still connect things, so the map will be usable. One commercial map maker has a number of 4WDs that continuously cross the country to monitor tracks and poi for changes (HEMMA .. they do good maps). Even so there maps are not ‘upto date’ all the time. Things change! The map will only be ‘upto date’ when the information is entered. Any time after that and it can be ‘out of date’. The expectation that the map MUST be upto date is unrealistic. It should be a reasonable representation for what was there … if there was a fuel supply at some place and it has gone.. there should be another fuel supply at some location close by. It is only where these features are a) vital and b) remote from similar thing that any map should be ‘upto date’. Only then is the requirement of being ‘upto date’ in any way important. Usually these kind of problems are provided by word of mouth at locations around the demise of the feature, people realize that maps in use may not have the current situation on them!

Comment from migurski on 26 September 2016 at 07:10

Hey Ben, thanks for this detailed response to my post!

I’ll just write a few words because I’m about to get on a plane. I think you’re pretty close to the point I am hoping to convey, but maybe coming at it from a different angle? I don’t believe OSM is irrelevant now (as evidenced by attendance at conferences like SOTM and the interest of the robot and crisis communities), but I think its leadership and the Foundation are stuck in a holding pattern that I trace back through the history of the project. Briefly, because the OSMF has been unable to adapt to the emerging needs of new members which I connect to CoC’s and improved license communication, those new members find reasons to limit their involvement in OSM and ultimately hesitate to buy in.

Money and all that it represents is a shorthand for this. The companies and consortiums who could conceivably provide this money find OSM unapproachable for the reasons I’ve described, and so they either get lightly involved or steer clear entirely. I’ve spoken about OSM with numerous people from large businesses and funding organizations who either muse about creating their own less-hostile alternatives, or dip their toes in tentatively before drawing them back. OSM’s unwillingness to open the door to participation on unfamiliar terms blocks the project from access to the kinds of funding you describe.

I understand that a number of people have taken offense at the term “craft”, and I am also aware that I’m making a disprovable prediction regarding project doom. I think it has been necessary to shake the tree enough to flush these topics into the open.

Also, BushmanK, I’d love to hear more about the proven fact that HOT involvement has not succeeded in sustaining the map!

Comment from ff5722 on 26 September 2016 at 08:53

While better in some areas, OSM isn’t on par, for the full range of uses, with maps from Google/Apple/etc.

Is that true though? I find other maps to lack in certain areas just as much. Google Maps is in China for example, where Baidu Maps and Tencent Maps have way more detail. For walking and cycling directions in Europe OSM is noticeable better than runner-up Google, with a lot more detail and a very up-to-date road network. Then you have the commercial aspect of other maps, it’s not clear based on which algorithm Google shows a certain shop at a higher zoom level, or they might adjust routing slightly to lead visitors through a shopping centre to boost sales.

And wasn’t the (original) idea behind OSM to just host map data? In theory, if there were contributors everywhere, we wouldn’t need an open source StreetView. For me at least, if I use StreetView (save for landmarks), it’s because the map is lacking and I need to get some information from the ‘raw data’.

I see a lot of opportunity for OSM with self-driving vehicles, when map data truly will become a massive business. When Google raised the fees for using their embedded maps, you saw the rise in use of Mapbox and other OSM sourced data. Big business can often kill open source projects, though of course there’s a risk in this as well.

Comment from BushmanK on 26 September 2016 at 15:09

@migurski,

I’d leave it to you since you love automation and so on: analyze a number of changesets in any area of Africa or Central America (Haiti, for example) within HOT/missing maps operation and outside of it. The fact, that HOT/missing maps operation is required for a certain area serves as the best evidence that there are not enough local mappers.

You can call for another operation and update it, but since HOT is emergency-related, you can’t just do it whenever you like (or you have to make up a reason for it). Between those operations, there are virtually zero edits. When were the last operation in Haiti and when - the one before it? Krymsk (the small town in Russia, where massive flood happened in 2012) is another example - Russian mappers were mobilized to map it, they did it really well, then it was abandoned since no more emergency situations occurred there. Keeping things up to date means being able to react to real-world changes more or less immediately, not working in a mapping raid style every time when flood, earthquake or something else happens. Missing Maps project is a bit different, but it has the same intermittent raid-style impact by its nature.

Do I have to explain further? I guess, no. Feel free to ignore this logic like you did before.

I’m perfectly aware, that it feels more important when you are not just editing some map, but also saving lives or something. Probably, it is possible to attract more people with this motivation. But it is impossible to call any region (especially, outside the not-so-developed countries) “an area of emergency”, which makes this approach limited by design. I’m not trying to say that HOT is bad or something, but it’s only better than nothing when it comes to a situation of emergency.

Comment from d1g on 27 September 2016 at 18:28

Nobody is talking about “craft mapping” or “armchair mapping” because these words are meaningless. But more, they are stupid and aggressive: http://www.openstreetmap.org/user/bdiscoe/diary/39538#comment36103

  1. place source=* tag on your changeset http://wiki.openstreetmap.org/wiki/Key:source
  2. please mention if your data sources are very old (years old or more); tell “survey” / “local knowledge” otherwise
  3. don’t bch about sources of other users like Harry Wood - http://wiki.openstreetmap.org/w/index.php?title=Armchair_mapping&action=history, don’t bch at your blog http://mike.teczno.com/notes/openstreetmap-at-a-crossroads.html

Comment from Harry Wood on 28 September 2016 at 23:14

@Mike:

a number of people have taken offense at the term “craft”

No the term “Craft” is fine. That’s why the term is co-opted and worn proudly as a T-shirt. The offensive bit was saying that people who map their own neighbourhood (hello? central idea of the project!) should be anything other than proud. That they will doom the project to irrelevance, while humanitarian mappers, and robot mappers will save it. It’s an offensive and highly divisive thing to say.

I think it has been necessary to shake the tree enough to flush these topics into the open.

Yes I’ve noticed in several follow-up comments from you, that you’ve climbed down pretty quickly. Ben here has done a great job of listing out your words which are over the top or simply incorrect. And here again now you’re saying the words were intended simply to “shake the tree”.

Flushing these topics into the open is good. The “Armchair Mapping” wiki page which d1g so politely refers to, is an example of my own attempt to flush out the topics. The page wasn’t easy to write. It’s walking a line between two strongly opposed viewpoints. There’s been a few contributions from others (it’s a wiki page after all) and some productive discussion on the talk page. Hopefully there’ll be more. It’s an example of an attempt to, not only flush out the issues, but help us reach agreement and bit more balance in the community, showing how opposing camps can accommodate each other. I hope more people read it (and read it a bit more carefully than d1g did)

The other way you could flush out the issue is by writing a divisive blog post which is offensive to many of the projects long-standing contributors. Well thanks. I’ve been paying close attention to these different groups and diverging opinions for a long time, and trying to bring more unity. I hope your blog will end up proving useful, as a way of flushing out these topics, but mostly at the moment it’s not feeling too useful.

Comment from d1g on 30 September 2016 at 03:10

No the term “Craft” is fine. That’s why the term is co-opted and worn proudly as a T-shirt. The offensive bit was saying that people who map their own neighbourhood (hello? central idea of the project!)

Maybe I think core idea is about getting 3D scanners and copters cheaper and legal, so we don’t have to go outside.

Maybe all or most of the tagging will done using AI.

Maybe we should have tags at wiki to be translated in 200 languages.

Promote well tested and productive/good idea - fine. Call yourself ordinalist/elitist because you are using good ideas - eh?.. OSM is not about this, use real social network if you really into it.

The “Armchair Mapping” wiki page which d1g so politely refers to, is an example of my own attempt to flush out the topics.

It was fine as draft/your opinion, until somebody decided to make guideline out of it. https://wiki.openstreetmap.org/w/index.php?title=Armchair_mapping&type=revision&diff=728729&oldid=728719

“Armchair mapping” was never widely discussed in Russian community but it was used by some as derogatory term here or there. Like it or not, “Armchair mapping” was used to abuse others and their methods and not to show what can be used in OSM

possible sources” can be used instead but you won’t be famous if Wikipedia won’t mention you as “coiner” of the two-word term. Hmm…

We don’t need to invent new derogatory words in English. We don’t need more derogatory terms in OSM.

I don’t think that creating a neologism in English language (a very ambitious move, indeed) would help us to describe what do WE have in OSM.

In OSM, WE have very few things to say about Geodesy, Engineering and other related fields who use Geodesy results only as intermediate data. 1cm is top-notch in Geodesy and bottom end in Metrology.

Sometimes you shouldn’t write an “Armchair mapping” article in your life.

“Armchair mapping” can be replaced with:

  • You haven’t seen one, don’t claim for it (Bing/knowledge/local knowledge/survey); don’t copy other maps and not accepted sources
  • Even if you seen one, It was long time ago, forget it; don’t fight others who seen it just recently,
  • Respect others who have more recent sources or more accurate results

and read it a bit more carefully than d1g did

You should work on your writing habits Harry and thank you for calling me a “patient reader” after I revised every single edit to Armchair mapping page but also all Internet results with this topic in two languages. For example, current http://www.openstreetmap.org/about page covers “current and local knowledge curated by many communities” way better and uses 10-50 times less words. Think about it next time.

Now when we have debates over techniques and “Craftmapping” (in 2016), we should state:

… look at the content AND CONSULT/INFORM LOCAL COMMUNITY.

For example; when we have automatic imagery classifiers/annotators, 80-90% verifiable data:

  • can be adapted by active community (interest + competence + free time)
  • it may kill interest of few mappers who just started work in their metro. Please wait with imports when “they are ready to fill otherwise empty spots” or “they agree to have % of errors in some dataset and work on it”

Login to leave a comment