A point consist of two double precision values. You store the same information, same size on disk and in RAM. But with a slightly different set of constraints.

Two separate columns can be NULL or not independently. A point can only be NULL as a whole:

• Why does the behavior of array syntax differ from '(?,?)' syntax when updating a point field and that field is NULL?

You can use geometric functions and operators for points. But you can easily construct a point from two numbers on the fly: point(x, y). In reverse, you can easily access coordinates of a point as double precision numbers: p[0], p[1] (0-based index!)

It's basically a matter of taste and style while you don't use the more sophisticated types geometry or geography provided by PostGis.

A point consist of two double precision values. You store the same information, same size on disk and in RAM. But with a slightly different set of constraints.

Two separate columns can be NULL or not independently. A point can only be NULL as a whole:

• Why does the behavior of array syntax differ from '(?,?)' syntax when updating a point field and that field is NULL?

You can use geometric functions and operators for points. But you can easily construct a point from two numbers on the fly: point(x, y). In reverse, you can easily access coordinates of a point as double precision numbers: p[0], p[1] (0-based index!)

It's basically a matter of taste and style while you don't use the more sophisticated types geometry or geography provided by PostGis.