[OSM-dev] WMSC - bachelor thesis

sorel johann johann.sorel at geomatys.fr
Wed Aug 17 21:06:31 BST 2011


I implemented clients for several 'tiled' services, TMS, Google Maps, 
WMTS, WMSC

If I may summarize my experience starting by the worst one.

4) Google Static Maps API (not talking about google maps api v3 which is 
only for web apps)
- limited to 1000 queries per day (by IP, so you can block your entire 
building connexion to this service in a few minutes of navigation if you 
have a big screen)
- max size 640*640
- single projection (EPSG:3857)
- no capabilities
- random responses after 1000 queries, sometimes succeed, sometimes not.

3) TMS
- minimaliste capabilities, lacks starting point coordinate, number of 
available pyramid depth, tile size, number of tiles on x/y
those simple missing informations makes tms unusable unless the user 
already knows them.
- most tms implicitly use EPSG:4326 in lon/lat axis order. while the 
true definition from EPSG authority is lat/lon order.
for professional tools this is one more nasty habit we found in plenty 
of projects. if you need lon/lat axis order use CRS:84.
- several tms implementation, not all working exactly the same way.

+ efficient
+ most used tiled system (excluding the fact they do not all work 
exactly the same way)

2) WMS-C

- overload WMS
- painfull to calculate the correct tile bbox (because of some server 
implementation not doing float rounding the same way)

+ provide all requiered informations to properly place tiles in any position
+ tiles can start anywhere on the projection
+ depth level can be irregular
+ tile size can be anything

1) WMTS

- I found no fully compliant server, all had some incoherent values. :(
(damned, is using an affine transform that hard ? ... looks like judging 
by all the mistakes I've seen)

+ provide all requiered informations to properly place tiles in any position
+ tiles can start anywhere on the projection on every depth level
+ depth level can be irregular
+ tile size can be anything


WMTS is not that complicate, it covers plenty of cases that's all.
I know most developers are not GIS developers, and do not understand 
most of
the problems of cartography,datums, elipsoid, transformation grids 
...and so on.
For them 3857 and 4326 are more then enough.
That's why WMTS might look like overcomplicated for them.
Yet it's not enough for needs above 'nice maps'.

my two cents.

johann



On 17/08/2011 19:07, Andy Allan wrote:
> On Wed, Aug 17, 2011 at 1:48 PM, sorel johann<johann.sorel at geomatys.fr>  wrote:
>
>> WMSC is obsolete, WMTS is the real OGC specificiation. osm uses something
>> which is not normalized, some kind of custom TMS service. several projects
>> out there us custom services, google maps, bing maps, yahoo maps ...
> Or alternatively, the OGC made a mess of the TMS specification and
> made it incompatible with the vast majority of TMS-like maps that
> existed at that time.
>
> Also, I've never heard of WMTS, but from what little I read[1], it's
> just an overcomplicated version of TMS.
>
> Gotta love OGC standards.
>
> Cheers,
> Andy
>
> [1] http://lists.osgeo.org/pipermail/tilecache/2009-September/002031.html




More information about the dev mailing list