On Apr 23, 2018, at 4:16 PM, Moriel Schottlender &lt;mschottlender(a)wikimedia.org&gt; wrote: Hi Strainu, Let me pitch in, at least from the technical point of view to some of your questions: On Mon, Apr 23, 2018 at 2:39 PM, Strainu &lt;strainu10(a)gmail.com <mailto:strainu10@gmail.com>> wrote: Just to clarify -- the change will not mean that all updates will be a matter of minutes. We are making the update from OSM to our systems more frequent, but the **best case** would be a minute. If you update a label just before we update, it will take minutes to update, but overall we are shortening the update length from every 48 hours to every 24 hours. That's the meaning of Joe's comment about how the updates will now take anywhere between minutes to 24 hours. It's still making it half as long! :) For the most part, <maplink> should also support these, as it calls the same server, but there might be slight differences in available features (like language override) for now. The main challenges to the displays and interaction that we're now busily fixing are more pronounced in mapframe displays, but the maplink system should follow the same stack, so it should also be updated when the system comes up, without having to do anything special. If a wiki enables mapframe, the feature will be available by default. This is more a policy question, I am not sure I can answer it, so I'll leave it to others. The fallback issue is actually pretty complex, and we've been trying to come up with the best case to display language in maps the best way possible; we've gone a bit back and forth with the algorithm as we examine the results, and it's likely that the consideration itself will continue to be slightly adjusted as we go. I will make sure that we document the technical behavior in a proper task on Phabricator, though. That said, to answer your question directly, we are, for the moment, getting rid of the "same script" behavior. As we examined maps, it actually makes no sense to display label names in a random language that just happens to share the same script. There are actually several issues we're concerned with, I am trying to summarize briefly: - It's really incorrect to assume that a speaker of language A can speak or read or make sense of names of a place in language B, even if it uses the same script. Place names aren't always the same in different languages. - It makes it much more unclear that labels in the requested language are missing - Some cases can be politically and culturally sensitive - We don't control the information in OSM, and some of it is missing a language tag (only tagged is the local language without specifying what language that actually is). The more aggressive we are in trying to add more fallbacks, the more we risk ignoring local names that are actually useful. So, to continue your example, it depends what area of the map you look at: - If you look for Romanian labels in China, it will show anything in Romanian, then it will look for anything that is latinized (suffix -Latin) in case there are transliterated labels, and then it will fall back to local language. You will likely see some labels in Romanian, some transliterated, and some in Chinese. - If you look for Romanian labels in Romania, you will likely get all of them (it will either get the direct language, or 'fall back' to local, which is Romanian) etc. It's important to note that in general, creating fallbacks for languages is a complex, and sometimes culturally sensitive issue, and that the fact we fetch the data from OSM means we don't have complete control over the way people added that data in. No matter what programmatic strategy we choose, there will be "holes" in the available data that will either need to be fixed in OSM by translating more labels, or by categorizing labels by their language better. Or, of course, by setting up more direct fallbacks for smaller languages. Thankfully, as you can probably see from the original email, Joe is being super thoughtful about any and all of those fallback considerations as we examine more and more maps, and we're really hoping to get as close as we can to the best experience possible, given the data and the scope of having maps for all over the world, displayed in so many languages. Hope that helped! Moriel _______________________________________________ Wikitech-l mailing list Wikitech-l(a)lists.wikimedia.org <mailto:Wikitech-l@lists.wikimedia.org> https://lists.wikimedia.org/mailman/listinfo/wikitech-l <https://lists.wikimedia.org/mailman/listinfo/wikitech-l>