[OSGeo-Standards] WMS-C and Capabilities

Jody Garnett jgarnett at refractions.net
Wed Aug 27 11:18:31 EDT 2008

Hi Satory -

You are correct - the mere act of tiling does not take too much. And in 
this case the start of the discussion was missing metadata.

The act of having the tiles make sense; we can draw them without much 
metadata. To go further (so I can present them to a user in a reasonable 
way) requires a bit more. The WMSC approach is a compromise between the 
WMS specification (which provides some description of the layers being 
drawn) and also makes these layers available in cached form as a TileSet.

Normally GIS applications can get away with "integration by 
visualization" (ie draw it all on the same screen and let the eyeball 
figure it out). uDig tries to do better; I build up the associations 
between a WMS Layer, a WFS FeatureType and now a WMSC TileSet - having 
all these options available to me allows the application to "be 
friendly" (it can choose the right tool for the job depending on zoom 
level for example).

Hope that all makes sense ... Please understand my feedback here is as a 
client trying to use the service. I understand it would be easier for a 
server to do less - the resulting information would be less useful to 
me. Normally server writers can get away with anything in this space 
because data is so valuable that clients will go through any tricks 
required to get something on screen. In this case I am opening up a 
discussion and with the help of the list modifying the wiki page above 
to be more clear.

I understand that I could also depend on a third party (say a catalog 
server?) to know about the association between tileset and layer. While 
that is an option; I would like to ensure the WMSC (or any tiling 
service) is self describing. At the very least those harvesting the tile 
servers can benefit from the additional information.


Satoru Takagi wrote:
> Hi all,
> There seems to be the considerable difference between the function 
> that is necessary for a tiling server and the function of the WMS server.
> The simplest tiling server can be achieved even by the simple file 
> system (For example, mere httpd without CGI, servlet, PHP..) which 
> stored tile files and metadata.
> Therefore, I think that the spacifications for tiling server should be 
> designed based on simpler mechanism (the basic URL based web) rather 
> expansion of the functions and its specifications of a dynamic server 
> such as WMS.
> Regards,
> Satoru Takagi

More information about the Standards mailing list