The trouble with imports (translated)

Posted by Richard on 9 November 2010 in English (English)

Paris mapper Pieren has written an excellent half angry, half regretful diary entry that sums up why unthinking imports are such a curse to OSM. For those that don't speak French, I'm posting a (slightly loose) translation here. Apologies for any errors of translation.

[start quote]

I sadly discovered today that the import of buildings in the 16th arrondissement [district of Paris] has been brutally completed by someone else, by means of the simple import of a file.

I'd been working on this area for several weeks and yes, that might seem too slow for some. But if I was progressing slowly, that’s because I was taking the time to check names and one-way streets on the ground, block by block; to clean up errors (disappeared names and POIs); to add extra information such as traffic lights, missing vélib [cycle hire] points, missing names, barriers, other POIs, addresses. Everything that a semi-automatic import doesn’t do (after all, I could have done the import myself in 15 minutes long ago).

I've already carried out this work for other arrondissements (smaller ones, true) over a long period. You only have to read my OSM diary to see the list. I was three-quarters of the way through working on this arrondissement, but here I'm stopping, disgusted. I shall leave the work of completing the missing information to others, so that the arrondissement may be consistent in a far future (if ever), as well as the other arrondissements of Paris.

I understand that many people work on each area, and it’s better like that. But a tiny bit of co-ordination and investigation, to check what others were doing, would seem to me to be a minimum before embarking on such an import.

[end quote]

Comment from imroy on 9 November 2010 at 16:40

Perhaps he should have been checking in pieces of his work as he verified they were correct, instead of waiting to add it in one big chunk. That way other people could see work was being done in the area (even finding his diary from user links in the history list) and made their own contributions on top. Instead, he's upset and OSM possibly suffers (at least in the short term).

This is a good demonstration of the differences between "the cathedral and the bazaar" styles of development (see Eric Raymond). In this case, Pieren was trying to use a more "cathedral" style of work in an environment set up for a "bazaar" style.

Comment from Pieren on 9 November 2010 at 17:14

You misunderstood. I didn't wait to add one big chunk. I imported and committed small chunks (quarters) during weeks, verifying that nothing was wrong and someone else came with its big chunk (not verified of course) stoping my progress. So it was impossible to not see my work.

Comment from imroy on 9 November 2010 at 19:29

ah, ok. Thanks for the clarification.

Did the other persons' import replace some of your work? Did it create duplicate nodes/ways?

Comment from Esperanza36 on 9 November 2010 at 23:34

I am the author of the "massive" import of 1/5 of the XVI district in Paris on Monday November 8th. First of all, I didn't delete any work from Pieren, I just added the last missing buildings of Paris. I didn't create duplicate nodes in this case. And after this "big issue", I have reverted all my work on this area, as Pieren wanted me to do in order to go on with his own work. I am very disappointed to be treated like somebody who mainly dashes off his imports.

In fact, one year ago, I performed the cleaning of West Bank import (dupe nodes, connections with Israeli road system and settlements) from Jumpstart with "Mikel Maron" and "talkat" users.

Therefore, I hope you will understand that my overall contributions to this project is not limited to this incident.

Comment from Pieren on 10 November 2010 at 10:18

I think Richard translated my diary because it's one example of how imports can disturb local contributions. I'm not against mass imports. I still believe that adding buildings footprint is very valuable for OSM and the semi-automatic tool we have in France to extract the shapes from the cadastre is very helpful and speeds up the process. It's just that contributors are working differently. Some prefer to import all buildings at once then fix issues or additional POI afterwards. And some (like me) prefere to progress by smaller chunks but more completed. It's a bit difficult when both ways try to work on the same location at the same time. I also understood that Esperenza is not working like this usually but made that way for some extra reasons. The matter is closed.

Comment from Govanus on 15 May 2013 at 18:41

When I make edits I hope to be correcting or adding to things from more knowledge usally in very small sesions though I am very aware that accurate geographic alignment of meny of my new lines and points is only as a close guess with the imagary behind (often bing), even when adding feature not readably visabile in the base pictures so if someone wanted to pull them into a better alignment with reality I'd be happy. I think that sometimes the tools like potlatch etc can be a little simplistic for working with previous efforts which isn't a total critisium as they are often easier to use. Though it is worth knowing this when you don't really know if someone has already spent a huge amount of effort with gps and other servay tools to align something accuratly or simply guessed it loosly or imported from a low resurlution source.

I think that stratigicly, as the map in places does become matured with a lot of acurate data in places, this is something that the main organisers will need to consider more to avoid looseing things already done or done more accuratly but with need for minor edits or just tag data updates.

I thinkk the way that OSM already records lots of change history is a useful aid to help begin making some tools or sub-apptools to address this type of co-ordination issues.

I know if I'd taken a data via a point-field with a calibrated lidar I'd be upset if someone just re-aligned it to the mistakes in a base image.

Login to leave a comment