I came across the traditional
ERROR: conversion between latin9 and LATIN1 is not supported
and read the thread latin1/latin9 in this list. The underlying
answer is essentially
http://www.postgresql.org/docs/8.1/static/multibyte.html#AEN22506
iso_8859_1_to_utf8 exists, and utf8_to_iso_8859_15 exists but not
iso_8859_1_to_8859_15, and "create conversion" as the documentation
page suggests might help.
But(!) I then wondered where the latin9 came from, and it seems to be
from:
case 1252:
if (PG_VERSION_GE(self, 7.2))
wenc = "latin9";
else
wenc = "latin1";
break;
in multibyte.c. (I'm looking at psqlodbc-08.01.0200 and running 8.2devel)
Why latin9? WIN1252 does exist (not sure when it was created - 8.1 same
as WIN1251?)
Cheers,
Patrick
