- From: Henning Schulzrinne <hgs@cs.columbia.edu>
- Date: Tue, 3 Mar 2009 21:25:40 -0500
- To: Andrei Popescu <andreip@google.com>
- Cc: Richard Barnes <rbarnes@bbn.com>, Alec Berntson <alecb@windows.microsoft.com>, public-geolocation@w3.org
> > > prd,pod,sts - can be coalesced with street (e.g. 'Carriage Drive > North') It can be, but makes dealing with street abbreviations much more cumbersome. I now have to parse Carriage Dr N and No Carriage Drive and decide which parts might be real names and which are just standard abbreviations for places. Maybe plausibly easy in English, a pain if multiple languages are involved. If I separate them, I can have a simple lookup table without worrying that 47 Platz Street (which exists in the lovely town on Toowoomba, Australia...) is not the concatenation of the German word for "place" with something else. > > hno - street number > hns - coalesced with street number (e.g. '52B') probably relatively harmless > > lmk, loc, flr, nam - can be combined into a separate field > ('details'?) These all have very different meanings. If you combine those, you suddenly have to understand the various-language versions of "floor", for example, to figure out which part is the floor number. That makes dealing with international addresses far more painful than needed. Everybody will be ordering them in their favorite order, so you can't just pick the first element. Similarly, distinguishing algorithmically and reliably between 'nam' and 'lmk' is between hard and impossible. Example: Mall of America Mallrat Store Which is which? Among other problems, this is makes it far more difficult to compare civic addresses with each other, to see if they are the same or "close enough". I may decide that I don't care about the floor, but certainly care about the building name. Catch-all terms are also likely to encourage dumping all kinds of non-address data into this. "Details" will likely attract free text like "The overpriced store with mall- shaped bed linens." > > postalcode - postalcode > > > So now we have 9 fields with understandable names. Alec's proposal had > 6 fields. Perhaps we can expand his proposal to accommodate the > missing fields as a compromise? > > What do you think? We are getting closer, I suppose. You haven't even touched the various international naming conventions, such as "Sub-Branch road name". We found out during our discussions that treating them as "just streets" was a recipe for confusion and required far more local cultural and language knowledge. I think it would be a good idea to have more memorable names for these, but that seems pretty trivial. I don't see much of a problem of understanding with labeling something "floor" or "landmark". Would it be too much to ask that you consider consulting the GEOPRIV archives, given that we have had these discussions at length there? I suspect you would expect the same if somebody were to enter into an area you consider your expertise. > > > Thanks, > Andrei > >
Received on Wednesday, 4 March 2009 02:26:24 UTC