On Thu, May 26, 2005 at 03:44:50PM -0400, Jan Wieck wrote:
> >For instance, I have a 7.4.x instance of postgres in SQL_ASCII. Could  
> >I subscribe to it with a node that has a default character set of  
> >UNICODE?
> 
> I'm not a unicode expert, but I would think as long as the encoding on 
> the subscriber is a true superset of the masters encoding with the 
> additional property that something unique on the master is also unique 
> in the subscribers encoding, it should work.

It will work _only if_ your SQL_ASCII database doesn't have
non-UTF-8 characters in it.

ASCII is a subset of UTF-8, in that anything in the 7 bit ASCII
charset is the bottom of the UTF-8 range.  The _problem_ comes if
someone put a non-ASCII character into your SQL_ASCII-encoded
database.  It'll go in as though there is no problem, and no
conversion or encoding changes will be performed.  So, if someone
sent the database a (for instance) ? (that's e-aigue for those of you
on a different encoding) encoded as ISO8859-1, the result will be
bad (??, if http://czyborra.com/utf/#UTF-8 is correct).  it gets
worse if you _actually_ have, say, ISO-8859-5 in your database
(because you can't tell, by looking at the data, whether it's 8859-1
or 8859-5, since they have the same binary values and are actually
only different if you select the right client side encoding.  If you
really want to be bored, ask me some time about internationalising
domain name labels).

A

-- 
Andrew Sullivan  | ajs at crankycanuck.ca
Information security isn't a technological problem.  It's an economics
problem.
		--Bruce Schneier