I was surprised this morning when I saw that a chunk of buildings had been imported in Victoria, BC. The changeset linked to the wiki plan and I then checked my email and saw this email chain.
The "local" (in this case the Canadian community) has not been sufficiently consulted. Looking back in the mailing list, there were some tangential discussions about some things related to this import (mostly without any final consensus), and then a single email stating that the import plan had been created and sent off to the imports list (https://lists.openstreetmap.org/pipermail/talk-ca/2018-November/008864.html). After that, nothing. I can't find any emails related to this import that were linked to both the imports and talk-ca list, nor are there any that bring back the results from the imports list for those who aren't following that. There also wasn't any notification that the import was actually going ahead. I'd consider all of this to be a major failure of the "Community Buy-in" section of the Import/Guidelines. For such a major import, we should be going above-and-beyond to make sure every possible aspect has been addressed adequately. The lack of confidence from the general OSM community as a result of the botched import in Ottawa and the ongoing dislike of our CANVEC imports means we need to treat this import with kid gloves. We should be striving to ensure that there's no reason why someone could look at this import and find faults with it. That may seem like a lofty goal, but we're talking about a building import for the second-largest country in the world. Once the administrative portions are dealt with and the community has been sufficiently consulted, the technical area needs to be looked at. Now that I've seen some of the data in action, there are various issues that need to be dealt with. Some that became immediately apparent on the data imported in Victoria, BC include: -A significant number of unsquare buildings (JOSM validator reports this as an Other/Building with an almost square angle). Of an estimated 935 buildings in this chunk, 692 have almost-square angles. Looking more closely and running the JOSM Orthogonalize tool on a sampling of buildings, I believe all of them have unsquare angles. This may be the result of rounding errors in data conversion and should be fixed in the source data before importing. -Inconsistent tagging (some houses are building=yes, some are building=detached) -A need for simplification (extra nodes in nearly-straight segments that are straight in reality) I'd suggest the following plan: 1. Update the tasking manager to indicate in clear terms that this import is on hold and no data should be imported at this time. Ideally, the tasking manager should be taken down entirely so no data can be imported. 2. Send a clear, unambiguous email to the talk-ca list indicating that this import is being planned and to solicit feedback. 3. Wait. THIS IS IMPORTANT! The community needs to be given time to see that this import is being planned, and to discuss the many aspects related to it. For such a major import, silence-as-tacit-acceptance doesn't fly. There are local communities out there that need to be brought in to the process. If necessary, figure out who the active contributors are in various jurisdictions and contact them directly. 4. Figure out the technical details. It's only after the import had already started that people are now talking about conflation and data quality. These need to be figured out, a plan documented, and the source data cleaned. Tags also need to be clarified. The current wiki plan gives almost no guidance about how to actually perform the import. 5. Only after all of the above has been figured out, let the community know that the import is actually going ahead. Come on, people. We can do a lot better than this, and definitely should. Let's make this a shining example of why imports can be a good thing for OSM, not provide fodder for those opposed to them. Andrew Lester Victoria, BC From: "John Whelan" <jwhelan0...@gmail.com> To: "Nate Wessel" <bike...@gmail.com> Cc: "talk-ca" <talk-ca@openstreetmap.org> Sent: Saturday, January 19, 2019 5:07:52 AM Subject: Re: [Talk-ca] 2020 building import wiki comment by Nate Wessel I'm saying when the matter was first raised on the import mailing list as a heads up I made reference to the existing Ottawa pilot and gave a link basically saying we would be following the same pattern. There was considerable discussion around the Ottawa import plan both on the import plan and talk-ca and the Ottawa import which I didn't draw up. Later there was a formal link to the data import plan. So two stages if you like. This is what we are thinking of doing and this is how we intend to proceed. Cheerio John Nate Wessel wrote on 2019-01-18 11:38 PM: John, I'm sorry to keep saying this, but I really do not think this is an acceptable import approval process. You're saying there was no wiki describing the plan when this went to the imports mailing list - only a link to a similar plan with related data. You did not follow the import guidelines and you need to go back and read that page line by line and follow the procedures that we have in place. https://wiki.openstreetmap.org/wiki/Import/Guidelines I'll go ahead and add a mention of this plan to the imports catalogue to get us started. I'll also add some sections to the wiki and try to leave some indication of where things can be better documented. You may think I'm quibbling over procedural details, but I think this process is really important. If we were talking about importing buildings in one neighborhood, I would look the other way, but this is all of Canada. This is a huge, huge import and we need to take the time to do things right, and especially to document the process so people can get involved that aren't already. Best, Nate Wessel Jack of all trades, Master of Geography, PhD candidate in Urban Planning NateWessel.com On 1/18/19 3:48 PM, John Whelan wrote: BQ_BEGIN The import mailing list was pointed to the correct page of the wiki. The initial post was to say this is what we were thinking of and there was a comment saying we needed to change the comment line. >There is no mention of this proposed import on the import catalogue The import process was reviewed by the person who set up the Ottawa import did we miss that step on the Ottawa import as well? Neither was it raised as a concern on the import mailing list. I think this is very minor and can be corrected. We learnt a fair bit on the Ottawa import and my expectation is since we are using experienced mappers to do the import conflation would be either handled by them or the building not imported. We aren't using new mappers in a mapathon here and with experienced mappers then I think you have to trust them. The world isn't perfect. Think in terms of service level. >There are 2X more nodes than needed to represent the building accurately. The problem with correcting this is you are introducing approximations. This will vary according to the source and this can be simplified or corrected once its in OSM. I think this is a different issue of a mechanical edit that needs to be considered separately. If we are concerned with database size then I suggest we change the instructions to say put the source comment on the change set rather than on the building outline. Cheerio John Nate Wessel wrote on 2019-01-18 3:06 PM: BQ_BEGIN John, You seem to be playing the long game with this data - it sounds like you've been working with this a lot longer than I have, and you've put in the time and effort to help make this actually-quite-incredible dataset available to us. I don't want to stop the import from happening - quite the opposite. I just want to make sure that the time is taken to do this right. OSM deserves that. Your (our) long awaited victory will be the sweeter for our patience now. There are several specific issues I see where the I's are not crossed, nor the t's dotted. I've mentioned several already, so I'll try to be brief (I really need to get back to working on my dissertation). 1) There was extremely limited discussion on the imports mailing list. The initial email did not make clear the scope of the project. I read the email and did not think twice at it, thinking it was entirely about Ottawa. The link in that email was actually to the Ottawa import, and not this one, which seems to have been only in draft at the time. https://lists.openstreetmap.org/pipermail/imports/2018-November/005812.html As such, this project has NOT been reviewed by the imports list, which is a requirement for proceeding with the import. 2) There is no mention of this proposed import on the import catalogue ( https://wiki.openstreetmap.org/wiki/Import/Catalogue ) which is required in the imports guidelines. I suspect many other guidelines have not been followed. 3) The wiki page describing the import is not adequate to assess the quality of the data or of the proposed import. See for example: https://wiki.openstreetmap.org/wiki/WikiProject_Canada/Canada_Stats_Canada_Building_Outlines_Import/Plan#Risks The import guidelines call for a description of how conflation will be handled. The fact that two of the major importers seem to have a substantial disagreement about how to handle existing data indicates this was not well discussed and I can see that it isn't well documented. 4) The buildings need to be simplified, quite a bit actually. Most buildings have multiple nodes representing straight lines. This bloats the database and makes things harder to edit by hand later. There are probably 2x more nodes than are needed to represent the data accurately, making it harder for editors and data consumers to work with down the road.This is a simple fix that will save countless hours later. ... I could go on, but I think this is plenty sufficient to justify pressing pause on all this. Again, I don't in any way want to disrespect the work that has gone into this effort already. We're all volunteers here and I know how much time this all takes. However. importing all/most of the buildings in Canada is a monstrously large task, which will have to dance around a lot of people's toes. We should expect this to take a really damn long time if we're going to do it right. We need to have the patience to learn from experience, from critique, and from the wisdom of the people who've learned from flawed imports in the past and have devised guidelines and processes so that we can have better experiences with this in the future. Nate Wessel Jack of all trades, Master of Geography, PhD candidate in Urban Planning NateWessel.com BQ_END BQ_END -- Sent from Postbox _______________________________________________ Talk-ca mailing list Talk-ca@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-ca
_______________________________________________ Talk-ca mailing list Talk-ca@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-ca
