[OSM-talk] Who interprets semicolon in tag values?

[Bryce Cogswell]

On Sep 23, 2013, at 10:45 PM, Jochen Topf <jochen at remote.org> wrote:

> On Mon, Sep 23, 2013 at 01:44:23PM -0700, Paul Norman wrote:
>>> From: Jochen Topf [mailto:jochen at remote.org]
>>> Sent: Monday, September 23, 2013 6:51 AM
>>> To: talk at openstreetmap.org
>>> Subject: Re: [OSM-talk] Who interprets semicolon in tag values?
>>> 
>>> I finally wrote down what I found out about semicolons in tag values and
>>> what I think about them. Turns out there isn't much software that
>>> interprets them and where it does, only in special cases. Thanks to all
>>> who replied and sorry that I haven't mentioned every case in my blog
>>> post.
>>> 
>>> http://blog.jochentopf.com/2013-09-23-semicolons-in-osm-tags.html
>> 
>> It's worth noting that when two objects are merged in iD it splits the 
>> tag's value into lists at semicolons, merges the lists from the two objects,
>> 
>> and then turns the resulting list into a semicolon separated list. See 
>> https://github.com/systemed/iD/issues/943. I believe iD is actually the 
>> only editor that attempts to support semicolon lists in tags. JOSM and 
>> P2 emit them, but don't interpret them.
>> 
>> This of course doesn't avoid contradictory tags like tiger:separated=yes;no 
>> but it at least avoids tiger:separated=yes;no;no;no;yes;no;yes;...
> 
> This would be better if semicolons always indicated sets of values, which,
> as I describe in my blog post, is the wrong assumption. For 'ref' tags this
> is probably the right behaviour, for about anything else it is wrong.
> 
> IMHO editors should either do something clever, if they know the specific
> tag and how it is used, or fall back to forcing the user to make a decision.

It would be great if support for semicolons was opt-in for tags. If a particular tag supports semicolons it should be documented along with the appropriate semantics (ordered vs. unordered, etc.) during merges, and with a hint in taginfo that editors could follow. Tags that don't opt-in should not get special treatment.