[OSM-dev] Some more notes on OSM vector-tiling

[Preet]

I'm interested in reading the paper, especially if it describes the
data format you guys are using. It'd also be cool if your website had
a running example so those of us without iPhones, etc can check it
out. There are a few libs that do geometry clipping very well,
especially for complicated area polygons such as coast line data, that
are available freely. I might have actually done something similar
(generated tiled vectors for osm's coastline) for an offline
application and I'm interested to see your process/methodology.

Preet

On Fri, May 25, 2012 at 9:07 AM, Tom MacWright <tom at macwright.org> wrote:
> Hi Sandor,
>
> Is any of this work open source, or have open specifications on the web?
> Statements like comparisons between filesizes of raster & vector data need
> to be cited.
>
> Tom
>
> On Fri, May 25, 2012 at 8:44 AM, Sandor Seres <sandors39 at gmail.com> wrote:
>>
>> Scale/zoom levels and tiling are essential for mapping servers, especially
>> if pretending on streaming transmission model. In case of a
>> vector/parametric data transmission service the scale levels’ generation and
>> the tiling of these, as a rule, is performed in quite a different way
>> compared to the traditional raster format based service (keep in mind that a
>> well constructed vector format may 20 – 40 TIMES be smaller than the
>> corresponding PNG raster format for the same content). We do an OSM vector
>> transmission based service for mobile apps (see www.fasterimaging.com).
>>
>> As someone properly emphasized a clipping is essential for any vector
>> tiling. But, while clipping of line-work objects (roads, streets, borders …)
>> is rather trivial, clipping of area objects is somewhat more complex and
>> complicated issue. Besides the clipping, some kind of area
>> reconstruction/restructuring has to be done (one container area may be
>> clipped into many parts, the same with the corresponding holes, the
>> restructuring has to decide which new holes are in which new areas, than the
>> issue of trivial tiles or empty tiles and tiles inside areas and so on).
>> Also, tiling inevitably results in a considerably larger data amount
>> compared to the original dataset. So, the question is – is it possible to
>> provide a server that combines the tiling’s efficiency and the data size at
>> certain, close to optimal, level. Fortunately, latest research and an
>> experimental version of such a server show that the answer is yes. The
>> experiments are performed on OSM vector data for Europe from some weeks ago
>> (Roughly 30 object classes/layers, 12 area classes like rivers, lakes,
>> forests, sea …, 12 line-work classes like roads, streets, paths, water lines
>> … and some point object classes. POIs and LBSs are overlays on such a base
>> map). The estimates also show that such a very simple server (no DB, no
>> caching …) is fully realistic with extraordinary performance (respond to
>> tens of thousands requests per second) and scalability (just make a new copy
>> as needed).
>>
>> A  white paper, describing in more details the above subject, is
>> available. Though in bullets format with many illustrations and with a
>> working title – Hybrid data format, multi tiling and a new server model.
>> Interested?
>>
>> Sandor
>> _______________________________________________
>> dev mailing list
>> dev at openstreetmap.org
>> http://lists.openstreetmap.org/listinfo/dev
>>
>
>
> _______________________________________________
> dev mailing list
> dev at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/dev
>