The one observation that's worth something is that you should make sure
that you're using an unambiguous data type for any primary key.

Thus, an integer column makes a "good" primary key; there's no risk of
ambiguity where two integers might be interpreted as the same value.

In contrast, using a timestamp or float or (I suppose) some of the GIS
types mightn't turn out so happily if it is possible for two distinct
values to be considered equal, and for that to be the case for a column
used as primary key.

My other observation is that in order for data to get replicated, it's
necessary for COPY to be able to serialize and deserialize the data.  I
don't imagine that's going to be much of a problem; broadly speaking, if
you can pg_dump the data, that means you can COPY it, and hence Slony will
be able to replicate it.  Conversely, if there's not a good serialization
function (e.g. - equivalent to "tostring()" which is common in lots of
languages), then Slony will have a problem, and it'll be the same problem
you'd have getting pg_dump to work.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.slony.info/pipermail/slony1-general/attachments/20140226/490e3f7f/attachment.htm