On 4-Oct-07, at 12:16 PM, Christopher Schmidt wrote: > On Thu, Oct 04, 2007 at 11:09:34AM -0500, Erik Uzureau wrote: >> There we go. What David's said here is what I was grasping for >> earlier here. >> >> If you use a singleTile layer that is gridded, *eventully* the >> user is >> going to pan the map such that the corners of the grid are in the >> viewbox. At that point, the only option is to load 4 tiles, at which >> point, we're back to gridded, in which case you might as well just >> use >> a normal gridded layer with a very big tilesize and a buffer:0 > > I think Paul is imagining a singleTile which matches the grid of a > *smaller* set of tiles -- so the gridded tile is 'snapped' to a corner > of a smaller grid. For example, you could imagine the grid existing at > 0.5 degree increments all the way across the world, and the tile would > snap to the closest one. The tile no longer has equal buffer space > on all > outside edges, but is more cachable. > > The size of a single grid square must be smaller than the smallest > area > that can outside the map at any given time, so when the new area is > requested, the tile 'jumps' to fill it, but positions itself based on > the grid instead of based on the map center.
right, the positioning based on grid rather than map center is the critical part since the requests are then more likely to be retrieved from the browser cache when returning to the same area. But the argument about wanting singleTile to be very dynamic is still a valid concern, to which I would say that in the current approach, client side caching becomes essentially a random event in which someone may get an old image and would not understand why. With a deliberately cacheable approach like snapping to the tile grid, we then have to think deliberately about how to avoid caching them and then actually get predictable behaviour. Paul > > Regards, > -- > Christopher Schmidt > MetaCarta > _______________________________________________ > Dev mailing list > [email protected] > http://openlayers.org/mailman/listinfo/dev +-----------------------------------------------------------------+ |Paul Spencer [EMAIL PROTECTED] | +-----------------------------------------------------------------+ |Chief Technology Officer | |DM Solutions Group Inc http://www.dmsolutions.ca/ | +-----------------------------------------------------------------+ _______________________________________________ Dev mailing list [email protected] http://openlayers.org/mailman/listinfo/dev
