Chris Holmes wrote:
So I don't think anyone ever claimed that GWC was officially supporting the WMS Tiling Client recommendation, and indeed it'd be impossible to be anything like 'compliant' as there is no test engine. And it's not even an OGC spec, it's an ad hoc spec made by a (great) crew of Open Source types.
Understood Chris; this is just an assumption I made when I saw the links. From my standpoint I need the information (that is missing) in order to make a good user experience. I have started up a discussion on standards@anonymised.com where I hope to sort out any questions/feedback.
The main driving reason to integrate GWC to GeoServer is to run the new KML super overlay stuff, which actually has little to do with the tiling client recommendation. The integration is pretty loose though, I don't think it'd be right to modify the main GeoServer WMS capabilities document with a vendor specific extension that most WMS clients don't support.
Um that is what Vendor Specific Capabilieis are for?
I could see having an alternate URL, like ?request=GetCapabilities&tiling=true to return a special caps document with the tiling stuff. But putting even that in while we're in RC status seems to be high risk, it's a new feature, not a bug. Could be good for 1.7.1 I was thinking of trying to push some other GWC integrations then, like having GeoServer's in memory tile cache also check the GWC cache.
I see your point; I was thinking that GWC was optional, but I am reminded that it is going to be included this round (primarily for the KML super overlay stuff).
The other way to skin the cat is to have GWC produce a valid capabilities document....
It could:
1. call the first server on its list (in this case geoserver)
2. fetch the capabilities document
3. splice in the vendor specific parameters
4. and forward the result to the client
If you really think inclusion of GWC in its present state would encourage clients that are not correct I think the thing to do would be to just remove GWC's capabilities document.
Actually it has already happened; if you check out uDig trunk you can see a GWC client that is displaying information to users in terms of TileSets rather than in terms of existing layers. The goal here is to present additional rendering options for some of the layers published by GeoServer.
It would have been a lot better to raise this objection earlier, I don't think anyone else was thinking that fully implementing the tiling client recommendation should be a requirement. It's still quite valuable as a caching engine for KML stuff, and for people to point openlayers at.
Understood Chris; I only just looked at the capabilities document today; and it never occurred to me that it would be like this. From my standpoint all the information is needed, so I did not guess that only the "vendor specific capabilities" section would be produced. The failure of the imagination is mine; and I should of reviewed the project some weeks ago.
The thing to do seems to be to get GWC to properly implement the capabilities document, to engage with that community, instead of hacking something in to GeoServer caps during a feature freeze.
I brought the limitation to the community the moment I realized and confirmed the limitation Chris. I could not of done more; now if there is a feature freeze then I am going to be out of luck. But I would like to talk to the community and understand what is going on first. I do feel like I am alone in the requirement to have a valid/complete capabilities file...
Jody