Sunday, March 01, 2009

OGC WMTS - open standard for fast web mapping

With the release of a proposed standard for tile-based web mapping, Web Map Tile Service (WMTS), the OGC is poised to provide an open alternative to proprietary web mapping services like Google Maps.

Worked on quietly by the gurus at CubeWerx, CREAF and the Autonomous University of Barcelona - the candidate WMTS Interface Standard is much like OGC's popular WMS, but it enables faster server performance.

As many know, OGC Web Map Server (WMS) has been criticized for being slow because it creates a new image for each request – rather than returning pre-generated tiles that provide an almost “instant” zoom and pan. WMS was designed this way because there were two goals behind the interface, interoperability and the ability to overlay many sources - and in this respect it's been very successful. A WMS client can overlay map layers from many sources in an arbitrary bounding box at an arbitrary scale with any number of styles. But this flexibility comes at a price - since a WMS server is required to generate each requested map image on the fly it's slower to respond than a tile map service.

But the OGC WMTS changes all this.

To improve performance, instead of creating a new image for each request, a WMTS returns small pre-generated images (e.g., PNG or JPEG) or reuses identical previous requests that follow a set of tile matrices. This proposed standard provides support for multiple architectural patterns - KVP, REST and SOAP.

WMTS is an evolution of Tile Map Service Specification by OSGeo and the Tiled WMS by NASA OnEarth. In addition, this is the first OGC standard to include a RESTful approach in addition to the usual OGC encodings.

Bottom Line - WMTS provides a natural way to evolve WMS services into a more constrained - but more scalable and faster service - so anyone can build services and applications that are fast, easy to use, and democratically accessible.

- Jeff


At 3/02/2009 09:41:00 AM, Anonymous Ed @L-3 said...

Hey Jeff,

Tiling must be hot now. See


At 3/02/2009 09:52:00 AM, Blogger Jeff Harrison said...

Yes, tiling is hot - but it shouldn't be done via proprietary interfaces on web mapping services like Google Maps. Open standards like the WMTS authored by CubeWerx should be used to return pre-generated images (e.g., PNG or JPEG) or reuse identical previous requests that follow a set of tile matrices. In this way services and online resources can be developed to share any location content available, wherever (or whoever) it comes from - and avoid proprietary geospatial service lock-in.

At 3/24/2009 05:35:00 PM, Blogger Brian said...

This standard looks pretty impractical if I'm correctly understanding that you have to support all of those different schemes, and the pixel resolutions don't go down as far as needed. 0.33m? I have higher res data than that.

At 3/26/2009 09:29:00 AM, OpenID Nuke said...

Not true. You can set the tile matrix to whatever scale denominator definitions you need. The well known sets are just a continence to meet existing tile standards (e.g. Google Maps).

At 3/26/2009 10:03:00 AM, Anonymous Keith Pomakis said...

Brian, you don't have to support all of the schemes. There are three interfaces defined - KVP, SOAP/XML and RESTful - and they are all optional. KVP and SOAP/XML interfaces were defined for compatibility with existing OGC projects and specifications, and the RESTful interface was defined due to popular demand. It's a bit unclear as to which of these interfaces will be the most popular, and it may even vary from community to community. Clients and servers are free to implement all three should they chose to be as maximally compatible as possible, but it certainly isn't necessary. My guess is that most clients and servers will choose to implement at least the RESTful interface, leaving the other two interfaces as relatively unimportant except for specific projects.

As for pixel resolutions, the WMTS specification is intentionally open-ended. WMTS servers can serve tiles of arbitrary scale/resolution. You can use it to map a circuit board if you've got the right coordinate system for it. The one place in the specification where specific scales come into play is in the concept of Well-Known Scale Sets. These are simply definitions of scale sets that are expected to be in common use. Servers are urged to support one of the Well-Known Scale Sets in order to increase interoperability with the rest of the WMTS world, but it's entirely optional. And compatibility with a Well-Known Scale Set doesn't mean that a server isn't allowed to extend the scales to finer or coarser levels either.

At 4/28/2009 03:07:00 PM, Blogger Joe said...

The comment that 'WMS is slow' is not true when a cache is created and an application calls on that cache for web-viewing. This is currently possible with proprietary/and non-proprietary apps that are used to create the cache, which can then be viewed in multiple map and globe viewers. What is WMTS bringing to the table that hasn't already been done?

The problem now is too many varied scales to cache from, and no agreement between agencies on how to handle this. What will WMTS do to help resolve this?

At 4/28/2009 04:34:00 PM, OpenID Nuke said...


You are dead wrong. Server side caching of a WMS raster map is a limited solution if that. Client side caching is hard to do, e.g. in case you want cashing of multiple zooms & regions.

Tiling is faster and allows a straight forward cashing on the client side. The ROA approach will also speed up things (access to prerendered tiles rather than dynamically preparing and serving a raster).

The WMTS is supposed to provide an interoperability standard to a successful map technique. It has other benefits such as enabling people to easily deploy their maps or aerial photographs.


Post a Comment

<< Home