[postgis-devel] spatial_ref_sys view

[Paul Ramsey]

I've been trying to figure out my objections to this, it worries my greatly but for amorphous reasons. I think, as I noted earlier, I cannot perceive the value of the improvement outweighing the complexity of the extra machinery. Some bullet point thoughts:

* Most users don't have custom entries, most users live in the space of long-existing EPSG codes. They get nothing out of this.
* Those users who do have custom entries, have largely figured out how to exist in the current regime. They use ranges that are out of the way of the system entries. Maybe they follow our documented recommendations (we have those, right?)
* Updates to EPSG definitions with new better measurements, etc, for existing SRS definitions will happen magically now anyways, since the AUTH_NAME/AUTH_SRID pair is what's used to map from an SRID to an actual projection transformation. Update your proj: get the new state-of-the-art for all existing systems.

So we're really down to just folks who absolutely need net-new EPSG definitions, but ... cannot figure out to load spatial_ref_sys.sql? More docs / better docs feels like a reasonable approach to this.

Basically, if you're sophisticated enough to use custom SRS definitions, you're sophisticated enough to read whatever best practices we want to provide to make sure those definitions easily survive an upgrade.

Something like that? I just cannot get comfortable with a complex re-working of a core piece that mostly doesn't make things any better for most users.

P