[OSM-talk] tagging and rendering

[elvin ibbotson]

I'm grouping a replies to several posts for this topic...


On 9 May 2008, at 19:13, Jeffrey Martin wrote:

> Typos in real words are easier to detect than a mistake in entering  
> a number.
>

In the scenario I was suggesting numbers would only replace words for  
type tags and users would never see the numbers but would just see  
words (in their own language) mapped to/from numbers in the database  
by the editor/viewer software. This somewhere between the ID numbers  
(set purely by software) and latitude/longitude (which users do not  
enter directly) and all the other tags, most of which (like name=)  
require user direct input.



> From: "Jeffrey Martin" <dogshed at gmail.com>
> Date: 9 May 2008 15:09:39 BDT
> To: "elvin ibbotson" <elvin.ibbotson at poco.org.uk>,  "OSM Talk"  
> <talk at openstreetmap.org>
> Subject: Re: [OSM-talk] tagging and rendering
>
> The rendering should be separate from the data. Marking a hiking trail
> as an autobahn so it will be a different color or be visible on higher
> zoom levels I think we all agree is wrong.
>
> Provided the data is correct, I don't see a problem with altering the
> way data is collected and recorded to make it easier for renderers,
> and those who program them and write the rendering rules.
>
> ----
>
> I can see the attraction to the use of numbers for the values of the
> highway tag. Having a new system that does not use terms that
> have other meanings can force people to think about the OSM
> definitions of the values. The UK centric terms have this effect
> for me. I have to think about what motorway means for the US
> or Korea in terms of the OSM definition because I have no competing
> definition of the term motorway in my mind. For me motorway
> only has an OSM definition.


I have today tagged a little country lane in my area as a railway  
line as well as highway=unclassified, as the free-from tagging system  
would seem to allow this and I wanted to see how it will be rendered  
by Mapnik and Osmarender.  I'm all for freedom but I think the type  
of a node or way is (like node ID and latitude/longitude) more  
fundamental than most tags which would retain user input and the  
potential to invent new tags.

> From: Jonathan Bennett <openstreetmap at jonno.cix.co.uk>
> Date: 9 May 2008 19:55:01 BDT
> To: talk at openstreetmap.org
> Subject: Re: [OSM-talk] tagging and rendering
>
>
> elvin ibbotson wrote:
>
>> Things humans read need to be human readable. The database should  
>> be  read by software and if it can be faster and more efficient  
>> using  numbers, numbers are what should be used.
>>
>
>
> The best way of proving this would be to come up with your own  
> version of the OSM server stack that used ID numbers internally,  
> while still outputting human-readable tag names. How long do you  
> think it would take you?

I don't think we want another server. I can already demonstrate it:
I am currently experimenting with binary data downloads for my mobile  
OSM viewer, mom. I need data for scales from 3 (just coastlines and  
country boundaries for enormous areas) to scale 15 (almost everything  
in a limited area) and until there is a binary API the data has to be  
sourced as XML then parsed to binary. The standard OSM API does not  
have any level-of-detail filtering so I am using XAPI. To get data  
for a particular scale I have to make several calls to the XAPI for  
each feature group (natural, highway, waterway, ...) in turn, and  
each call takes quite a bit of setting up in the code. If, for  
example, the feature types were structured using a numerical system  
such that  so that all natural features began with 0, all highways  
with 1, etc, but everything needed at scales smaller than 5 ended  
with numbers smaller than 3 (eg. coastlines: 01; trunk roads: 12) I  
could make a simple call for features less than *3.



> From: "Martijn van Oosterhout" <kleptog at gmail.com>
> Date: 9 May 2008 19:51:18 BDT
> To: "Jeffrey Martin" <dogshed at gmail.com>
> Cc: OSM Talk <talk at openstreetmap.org>
> Subject: Re: [OSM-talk] street traits
>
>
> On Fri, May 9, 2008 at 8:30 PM, Jeffrey Martin <dogshed at gmail.com>  
> wrote:
>
>> A name for each kind of road in a person's country could be set up  
>> as an
>> editor feature. I select
>> "mountain road 2" from my list and it fills in the number of  
>> lanes, lane
>> size, shoulder size, etc.
>> for me.
>>
>
> Strangly enough, JOSM supports this already.
>
>
>> Another option might be to have some kind of bot that fills in  
>> specific data
>> based on country
>> specific highway tags.
>>
>
> I thin you're missing something though. Just because it says
> highway=motorway doesn't mean it looks identical everywhere. It means
> what a motorway is in the country its located in. Just determine which
> types of roads there are (there are about 7 usually, no matter what
> the country) and then map those to the existing highway tags. All
> done.
>
> If you want to add stuff like lanes/etc go ahead, but for the basics
> you don't need it.


A new thread but more evidence that feature type tagging is  
fundamental and may need rethinking. 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20080512/9149ca3e/attachment.html>