Hi, On 19.02.20 13:14, Christoph Hormann wrote: > the document then almost exclusively presents > supposed exceptions from the attribution requirement of the ODbL.
I've just read the document for the first time this morning, so I don't have the context of prior discussions and I think your wholesale dismissal isn't justified. > Or in other words: It is the preemptive surrender of the OSMF in front > of massive corporate interests. I think that the document has quite a few bits that do not exactly sound like a surrender, for example: * "compliance with these guidelines today does not mean that we will not propose or ask for different attribution in the future if it promotes our shared goals" - a good assertion of our rights. * "Except for small maps or multiple data sources, as described below, attribution must be visible without requiring the user to click on an icon or similar interaction." - Your critique focuses on the exceptions, but saying clearly that an "(i)" is *generally* not sufficient is a good and necessary step. On the whole, I find that the document does a good job at fleshing out the "reasonably calculated to make ... aware" from the ODbL. > Not to mention the most blatant attempts at sneaking corporate wishlist > items into the guideline are all still there - like the 10000 m^2 map > area limit that has been conjured out of thin air True, this is a bit strange, it would have to be replaced by "an area of up to 1,000 inhabitants" as per the "Substantial" guideline - though I don't find the difference outrageous, in fact the 10.000m² will only be *friendlier* towards non-attribution than the "1.000 inhabitatants" in densely populated urban areas. I guess that 100m x 100m is simply easier to check than whether the area has 950 or 1050 people living there! > the > section on machine learning models which is completely out of place in > an attribution guideline and which indicates that some corporate data > user wants this kind of "blank check" really badly. I agree that the attribution guideline should not be the place where we discuss what does and what does not constitute a derivative database. Perhaps the section should be removed altogether. In my opinion, if you train your AI black box with OSM data then everything that comes out of your AI black box later is a derived work and must come under the ODbL. I welcome the acknowledgement about "over-trained" systems creating ODbL output, but I think it doesn't go nearly far enough. Everyone and their dog are crawling over OSM with their AI stuff in order to build machines that can "map automatically", but essentially it is our brainpower that allows them to train their machines so it's our license. I acknowledge Kathleen Lu's recent remark about the ODbL being very clear on a derived product having to "contain" OSM in some way which would not be the case here; but I think this calls for working on ODbL 1.1 to rectify the issue, rather than sitting back and saying "uh, guess there's nothing we can do then". Bye Frederik -- Frederik Ramm ## eMail frede...@remote.org ## N49°00'09" E008°23'33" _______________________________________________ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
