MacRoman character set

Hi,

I've got an app using the libpq interface that uses the mac roman character set (natively, and I don't think I can change it). I'm using the 7.2.x releases of pg right now. The clients are all running OSX, some of the servers are on OSX, some on debian linux. (for the hackers, a MacRoman - unicode conversion table is here: http://www.unicode.org/Public/MAPPINGS/VENDORS/APPLE/ROMAN.TXT ) 

I would like to transparently encode/decode to SQL_ASCII all connections from this client app.  I don't necessarily want to encode/decode all connections from this ip address, as backup and psql are both run on the machine as well.  There are four possible locations that I can think of to insert the conversion:

1) do it in my app. - violates the transparency desire
2) do it in the glue between my ap and libpq - being investigated
3) create a new character set in the backend and set the client encoding.
4) use the --enable-recode option to make the change on an ip by ip basis - this complicates backups.

Of these, I think #3 is the cleanest, but it's also the one that I have no idea as to the scope. How hard is it to add another character set as a conversion target, but not one allowed as a backend storage format (sort of like the windows character set that doesn't have a unix code page?)  If it's not that hard, is this something that could be applied to the 7.2 series or would I have to upgrade to current? 

eric