Reprojecting with Gdalwarp vs native MapServer

[Gregor Mosheh]

> I have noticed that reprojecting layers on the fly in MapServer is much
> slower than creating a new file with gdalwarp and using that one in
> MapServer.

Any time that data needs to be reprojected on the fly, it will be a lot
slower than spending the time to reproject the entire dataset beforehand.

Consider: You make a request for a certain bbox. Mapserver needs to
reproject the entire dataset, then test each feature for spatial
intersection with that requested region. With every pan, every zoom, every
query. Yuck.

Doing the reprojection beforehand saves a lot of work at runtime, which
translates directly into faster performance for the user. That'll always
be the case, regardless of what reprojection engine is used or what CPU
you have.