[Talk-GB] Peer verification (was: Liam123 again)
On Fri, Aug 07, 2009 at 01:11:02PM +0100, Nick Barnes wrote: To my mind, nobody ought to be able to edit live map data unless: 1 - They have uploaded n tracks, 2 - They have had m edits approved by a moderator 3 - They are vouched for by somebody who has made many many edits (insert 'and' or 'or' or 'and/or' as appropriate) It has already been said, but I think raising the barrier to contribution is the wrong way to go. Instead, I’d like to see a way of saying someone has verified the data without changing it. This has been talked about by others before, and I think every State of the Map conference has had presentations on the subject. Multiple users should be able to indicate that they have verified the data. Subsequent edits would be unverified until somebody says they have verified it. Map users wanting a probably higher (but still not guaranteed) quality of data can then get a verified set, and they can choose how much verification they want, e.g. data that has been verified by at least two others. This doesn’t stop rogue editors creating extra accounts to verify the data. We can, however, limit who is able to verify the data without raising the barrier to contribution of (unverified) map data, although I’m not sure that’s necessary either. Taking this further would be to build up a trust network, and allow people to say “I trust edits made or verified by user X” and “I trust edits made by someone who has made more than N edits verified by someone I trust”. You can do this without digital signatures, but I would also like to see support for OpenPGP signed changesets and verifications, combining trust in people to correctly edit/verify data with trust that it was actually edited/verified by that person. The result maybe that if you apply all of the above, you get a very small amount of map data, and it may even be pretty worthless because there are so many gaps, but you get to choose how you deal with that. You can get a small amount of highly trusted data that may have gaps that make it less usable, through to a large amount of untrusted data that may have critical errors, or combine the two and possibly indicate where data is untrusted. Simon -- A complex system that works is invariably found to have evolved from a simple system that works.—John Gall signature.asc Description: Digital signature ___ Talk-GB mailing list Talk-GB@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] Peer verification (was: Liam123 again)
On Sat, Aug 8, 2009 at 11:11 AM, Simon Wardsi...@bleah.co.uk wrote: It has already been said, but I think raising the barrier to contribution is the wrong way to go. Instead, I’d like to see a way of saying someone has verified the data without changing it. http://wiki.openstreetmap.org/wiki/Proposed_features/Checked_by ___ Talk-GB mailing list Talk-GB@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] Peer verification (was: Liam123 again)
On Sat, Aug 08, 2009 at 01:55:30PM +0100, OJ W wrote: Instead, I’d like to see a way of saying someone has verified the data without changing it. http://wiki.openstreetmap.org/wiki/Proposed_features/Checked_by I’ve seen that proposal before, and it feels a little “icky”. The comment by Kumakyoo highlights one of the problems with it, and your solution of cross‐referencing with the history would work, but that sounds like a lot of work for what should be a simple task. I also feel that only a user should be able to add their own name to the list. You could check that against the username on the changeset, but again it seems like making more work than necessary. It also leaves with the possibility that the tag easily builds up over time with usernames who have checked previous edits being left in after new edits have been made. Keeping the tag clean means maintenance work, and changes would ideally be made to either the server side code to treat this tag specially or for every editor to additionally maintain the tag. Although it may seem unlikely to get so many people verifying at this time, based on ~10 characters per username, the maximum number of characters allowed for tag values in the current API (255 AFAIK) is reached with the 26th user to add; it doesn’t scale very well. On a data level, the checked_by tag isn’t really relevent to the feature in the map data. It is data about the particular state the data was in at the time, more along the lines of the username and timestamp for an edit. I haven’t thought much about how it would be implemented, but I think you want at least the following easily queryable, especially if you’re going to be pulling data out based on rules involving the people who verified the data and how many verified it: element id, element version, checked by user id, date checked. Simon -- A complex system that works is invariably found to have evolved from a simple system that works.—John Gall signature.asc Description: Digital signature ___ Talk-GB mailing list Talk-GB@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-gb
Re: [Talk-GB] Liam123 again
On Sat, Aug 8, 2009 at 4:34 AM, Jeffrey Martindogs...@gmail.com wrote: Maybe we want different policies for different areas and different kinds of data. For example once all the roads are mapped we freeze the roads, but we allow free changing of street names until they reach a freeze point. Here in Korea I just want data and the more the better. In downtown London I would assume all the roads can be frozen except for major construction. That would be a very bad assumption, as every single one of our continuing London mapping parties shows. I'm constantly moving and renaming roads as the data we get becomes more precise. Nevermind the more technical issues such as adding and connecting roads to existing roads, or foot paths, or cyclepaths, or any of the other stuff which might not be considered mapping the roads but still requires editing them. Dave ___ Talk-GB mailing list Talk-GB@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-gb