Comparison for seeding optimization

[SEWilco]

For datasets where changes are concentrated in one area, first seeding areas which have changed may help seed regions which are getting the most viewers. Due to processing overhead this probably should be a configuration option.

For example, a national radar map may have only a few regions with data, and viewers are likely to be looking at those regions. The regions which have changed would also be those which should be seeded first.

Comparison of a new tile with its predecessor can identify large regions which have significant changes. The tiles which had changed, or by an arbitrary metric have the most changes, could get priority for seeding tiles at the next depth. I don't know if there are binary or perceptual image comparison tools in the presently used libraries.

This may be related to the ticket for depth-first seeding. http://geowebcache.org/trac/ticket/5

[arneke]

These are very abstract requirements, maybe more suitable for a developer mailinglist discussion than a ticket ?

We want to do GeoRSS in the long run, or have the backend server send truncate requests to invalidate a set of tiles. I don't think comparing images is feasible or a very efficient way of dealing with this problem.

[arneke]

Cleaning up issues prior to 1.0