On 15/01/2016, Kieron Thwaites <kieron.thwai...@gmail.com> wrote: > I completely endorse the removal of any and all *_N tags.
If so, you've got a serious amount of work comming up just to figure how to say the same thing using different tags. Semicolons and various namespaced schemes sometimes do the trick, but outright banning *_N for the sake of (what ?) would cause a lot of headaches. On 15/01/2016, Martin Koppenhoefer <dieterdre...@gmail.com> wrote: > 2016-01-15 15:26 GMT+01:00 moltonel 3x Combo <molto...@gmail.com>: > also shop_1 tags are created that way. I wonder why you would want to add > these tags on purpose. E.g. for shops the values indicate a type of shop. A > bookstore that also sells music cds? Still a bookstore IMHO. A music store > that also sells some books? Still a music store. Now, a store that sells > both, music cds and books, and maybe dvds and posters, maybe still a > bookstore, maybe a new category like a media store, not sure, would have to > decide on the particular case. In a majority of cases you still can decide > on one of the types and that it is. A green grocer that also sells > toothpaste, detergents, toys and cheese? That likely not a green grocer any > more, that's either a convenience store or a supermarket. > What I want to say: a shop that is 2 shops in one might well be neither of > these shop types, but a new typology. > In other cases, there is a well defined shop/restaurant/bar/kiosk/... which > also offers some odd service or product you might find in some but not all > of this kind of business, and which you consider so interesting that you > want to map the presence of it. Examples might be tobacco/cigarettes, > icecream, particular soft drinks (club_mate comes to mind [1]), public > transport tickets, fresh milk, etc. Yes, "shop=foo shop_1=bar" is not great. In the same way that "shop=foo;bar" is not great. I'm all for either finding a new fitting value or for tagging the main value plus a subtag, or for using two objects. But often mappers are not knowledgable enough or do not have the time to fuss about this, and instead "simply want to express that this is a deli and an optician". > My solution for this is sells:foo=yes(/no/etc.) > Obviously we wouldn't want people to tag the whole assortment of a > supermarket like this, but due to the amount of work the risk is low. > People will likely just tag the things they are particularly interested in, > and that are not automatically thought of being available generally. So far > the list is small ;-) [2] Yes, ns:foo=boolean is a good alternate scheme, when you can use it. > For names, the solution should be to use well defined name key variations, > there's a whole lot of them [3], and introducing just another infinite > amount of indexed ones seems completely unnecessary. The problem is that all those name key variations carry semantic meaning. A loc_name isn't the same thing as an alt_name which isn't the same thing as an old_name. You can't shuffle all your names into random foo_name keys, it has to make sense. And as soon as you've got more than one name, you've got a problem. Which is solved very nicely by _N. To get back to my http://www.openstreetmap.org/relation/5257865 example, I've got 3 names to tag. One of them distinguishes itself by also appearing on an out-of-copyright map, the other two are at the exact same level with each other and, as far as I'm concerned, pretty much at the same level as the first one. I can't fit them into "loc" or "old" or "whatever" categories, to the best of my knowledge they are just "other names". Which is solved very nicely by putting them in name_1 and name_2. Having multiple values for one tag is awkward in OSM, so we always try to find clever ways around it. Sometimes that clever trick is the right thing to do, sometimes we really just need a way to tag multiple values. Semicolons and _N are two ways to do that, each with their pros and cons (I don't think it likely or desirable to deprecate one for the other). They are sometimes usefull, please leave them be an accept them as another tool in your mapper's toolbox. Changing the topic a little bit, I'd like to comment on alt_name vs name_1 va alt_name_1. To me name_1 and alt_name are exact synonyms, I don't see a semantic difference (as opposed to loc_name for example). Therefore, if you've only got two names to tag you can use either. But once you've got at least three you'll need to use either alt_name_N or name_N. I find the concept of alt_name_N silly and would rather use name_N, but I've seen some people disagree, For what it's worth, alt_name_N is much less frequent in the db than name_N, but Nominatim supports only alt_name_N. Any opinions on that issue (other than "remove all *_N" :p) ? _______________________________________________ Tagging mailing list Tagging@openstreetmap.org https://lists.openstreetmap.org/listinfo/tagging
