Autodetect image format for a WMS server

OpenLayers will never ask for getcapabilities by default -- the majority of WMS-based OpenLayers applications use remote WMS servers with no proxy available, and using GetCapabilities would require working around the Same origin policy in some way. However, making it easier to request this type of behavior is not beyond the scope of OpenLayers. I can think of two different implementations:

1. option on the layer to ask for getcaps for image format determination before sending request. This would mean that the layer could not be loaded until after this request was complete: I think the way to solve this (at least for non-baselayers, I don't know if it would work for baselayers) would be to set the visibility to false before attempting to load the getcaps, then setting it to true again after the information is fetched and assigned. 2. OpenLayers.Layer.WMS convenience function for passing an OnlineService and a layername, and getting back the constructed WMS layer. THis could include support for multiple URLs (which I'm told WMS can describe).

However, I don't think that either of these is neccesarily the right way to go: instead, I think we just want to write a WMS capabilities document parser, and leave the loading of this document and the creation of the layers from it up to the application. These convenience methods are going to be silly: OpenLayers layers must be configured by an application developer, and the application developer has to pull the information about the layer configuration from a datasource which has to include at least two pieces of information: url, and layername. Storing the format that the layer prefers along with this datasource does not seem sufficiently arduous that complicating the process by which layers are constructed in any way makes sense.

In addition to the fact that this isn't solving a technical problem, but a social one, there's also the fact that parsing getcaps means additional load on the server which is unnecessary: Due to OpenLayers not having the ability to serialize data for later use, there's no way to parse the document once and reuse it again later. Instead, the document would have to be parsed each time the map was accessed: this is only going to cause unnecessary and visible delay to the user who has to wait for the WMS request to return before the map can start displaying tiles. I can't think of an application where this makes more sense than simply configuring the layer correctly when it is created.

I could be wrong on this: I'd like to understand what I'm missing about how additional requests to the server are designed to improve the behavior of OpenLayers, or how the solutions I've described above don't adequately solve the problem.

Since WMS Capabilities format is already targeted for 2.7, I believe, and I don't believe that pursuing these additions to the WMS layer is likely to be actively developed in the near future, I'm putting this in the future milestone. If someone wants to implement this, I suggest taking it to the mailing list first, as it is likely that at least some discussion of a use case would need to be worked out before code was written.