As is probably common I use the BBOX returned during a refresh of a View-Based refresh to return some results to the user. (in my case a bunch of placemarks or a grid, but that's beside the point)

This works well when the view is directly overhead as there is very little wastage and the bbox represents the region of interest quite well. However as soon as the view is tilted it becomes less useful as lots of the bbox is wasted in encompassing the distant area toward the horizon.

While this is visible there's not much point returning high resolution data for it as 1) is barely visible and 2) wasting bandwidth which could be better repositioned to show the foreground in higher res...

So my first question is has anybody got a nice solution to this? #

To combat this I have a solution that seems to work quite well...
Basically I capture the bbox and the lookat lat/long, and if the lat/long is outside the central area of the bbox (hence got some tilt going on), recalculate a new bbox square sized based on the distance of the lat/long to the nearest edge. As a small tweak I move this square forward toward the center of the bbox to place it more in the users foreground (less wastage behind the view)

I realise my explanation is probably garbled so a demo might help, luckily I have one here, this is a copy of my UTM grid-lines layer with this tweak. Try this out with when tilted and hopefully the grid should be high-resolution in the foreground area, without getting diluted extending into the distance. (compare with unmodified here - both these links should open the network link into your Temp places)


And so my second question, is does this seem to work well, i.e. does it seem logical as a user? I think so, so I plan to implement this** on each network link I control


Source code of the bbox recalculation snippet here, both as a RFC and a help for others.

# I know could probably use a combination of the lookat vars, to calculate a bbox, but frankly working out the size of the bbox is a little beyond me. (I can think up unconvential solutions like the above, but not do the 3d geometry )

** well a slight variation for the placemark links which don't need to be constrained to a square 'area'.