Further to my GIS PointOnSurface (POS) email, which didn't garner any replies (no GIS programmers here?), I managed to "solve" it by doing both POS and Centroid and making sure I use the point returned by the function that actually gives me a point that is actually ST_Within() the polygon(s) (according to MariaDB). ST_Centroid actually turned out to be much better, only failing on one state... guesses? ... Hawaii. Ya, guess the centroid of those polygons is somewhere in the pacific. This still isn't a real solution because a centroid AND POS could both not give me a mysql-happy point within the state, but since these things almost never change, and weird states are already rare, it should be good enough for now. And when Mysql/Maria give me some better functions, like other systems do, I can just switch to using those. Here's the states and whether they pass the centroid ("c") and POS ("pos") tests or not. Note how many fail POS. +------+------+------+ | abbr | c | pos | +------+------+------+ | AL | 1 | 1 | | AK | 1 | 1 | | AB | 1 | 1 | | AZ | 1 | 0 | | AR | 1 | 1 | | BC | 1 | 1 | | CA | 1 | 1 | | CO | 1 | 1 | | CT | 1 | 1 | | DE | 1 | 1 | | DC | 1 | 1 | | FL | 1 | 1 | | GA | 1 | 1 | | HI | 0 | 1 | | ID | 1 | 1 | | IL | 1 | 1 | | IN | 1 | 1 | | IA | 1 | 1 | | KS | 1 | 1 | | KY | 1 | 0 | | LA | 1 | 1 | | ME | 1 | 1 | | MB | 1 | 1 | | MD | 1 | 1 | | MA | 1 | 0 | | MI | 1 | 1 | | MN | 1 | 0 | | MS | 1 | 1 | | MO | 1 | 1 | | MT | 1 | 1 | | NE | 1 | 0 | | NV | 1 | 0 | | NB | 1 | 1 | | NH | 1 | 1 | | NJ | 1 | 1 | | NM | 1 | 1 | | NY | 1 | 1 | | NL | 1 | 1 | | NC | 1 | 1 | | ND | 1 | 1 | | NT | 1 | 1 | | NS | 1 | 1 | | NU | 1 | 1 | | OH | 1 | 1 | | OK | 1 | 1 | | ON | 1 | 1 | | OR | 1 | 0 | | PA | 1 | 1 | | PE | 1 | 1 | | PR | 1 | 1 | | QC | 1 | 1 | | RI | 1 | 1 | | SK | 1 | 1 | | SC | 1 | 1 | | SD | 1 | 1 | | TN | 1 | 1 | | TX | 1 | 0 | | UT | 1 | 1 | | VT | 1 | 1 | | VA | 1 | 1 | | WA | 1 | 1 | | WV | 1 | 1 | | WI | 1 | 0 | | WY | 1 | 1 | | YT | 1 | 1 | +------+------+------+