Add Tilecache support to Mapserver

[woodbri]

Assefa created a patch to mapserver that adds WMS-C Tilecache support to mapserver. This is on the enhancement list for 6.0 ​http://trac.osgeo.org/mapserver/wiki/60ReleasePlan and this ticket to to collect patches and discussions for this feature until a formal RFC is written.

Assefa's original code worked against 5.4.2 on Windows and has been included in MS4W. I have ported the patch to Linux and to version 5.6.1, 5.6.4 and 5.6.5.

[woodbri]

In addition to WMS-C tile requests, I have a need to request tiles using VirtualEarth quadkey, similar to mode=tile tilemode=ve. For more inforamtion on VE tiles read:

​http://msdn.microsoft.com/en-us/library/bb259689.aspx

Assefa hinted that it we might consider keeping the same WMS-C tile storage, but mapping the quadkey ID to an appropriate tile. I think this would be a very good idea to maximize the reusability of the existing code while supporting multiple request protocols.

[assefa]

I believe It should be possible to do the conversion. From what I can see you can for example refer to VETMS.py inside tilecache.org for conversion code: ​http://svn.tilecache.org/trunk/tilecache/TileCache/Services/VETMS.py ). You also need to make sure that when paring the URL, you look for a "tile" parameter and assume that if it exists, It is a VE type of request. (As opposed to WMS request that we support now).

[dmorissette]

Maybe I'm missing something, but should we not use explicit "mode=vetile" to be in line with RFC-43 (Gmaps tile support) instead of trying to automagically detect "tile="in the URL and risk causing side-effects?

[woodbri]

I don't care what arguments trigger the behavior. I also thought of using that, but you still need to pass in the quadkey to fetch the appropriate tile. I also have not thought about combining the gmaps tile support with this or if there is code that could be reused. I'm guessing there are a bunch of common utilities for converting between x/y/z, quadkey, etc.

It turns out my client will always be precaching the tiles which I already can do without this, so I'm probably not going to make an progress on this for now. I think this needs an RFC that Assefa was going to put together at some point if he has time.

[dmorissette]

Actually, I forgot to add that your suggestion of "mode=tile&tilemode=ve" would be just as good, I mostly meant that the interface should try to be consistent with what was done in RFC-43 from the user's point of view, and also cast a vote in favor of explicit mode=... args instead of trying to automagically detect the type of request based on the presence of the tile arg, which could cause some problems in the long run (see ticket #2531 for an example of what auto-detection can lead to).

[assefa]

Daniel, I agree on using/parsing explicit set of parameters. I did not really thought it through in my first comment; I was more focused on the fact the quadtree keys could be converted to tiles.

[woodbri]

I'm closing this as wontfix because this request is superseded by mod_geocache project that is a better solution to this problem and slated to merge with mapserver project this year.