On Fri, Oct 3, 2008 at 9:28 AM, Christopher Browne
<cbbrowne at ca.afilias.info> wrote:
> The way I would prefer to make this change would be a little bit more
> complex than the above.  The BINARY form is documented to be "less
> portable," as follows:
>
>   "The BINARY key word causes all data to be stored/read as binary
>   format rather than as text. It is somewhat faster than the normal
>   text mode, but a binary-format file is less portable across machine
>   architectures and PostgreSQL versions."
>
> Thus, my inclination would be to add a configuration parameter for
> this, which allows changing between BINARY and "not BINARY" at
> runtime, as opposed to picking one policy or the other at compile
> time.

The best solution would be to dynamically figure out when it's OK to
use binary copy and when it's not.

Not sure how to best figure that out, though. :-)

Is there a way to determine the version of Pg and architecture the
other node is running? If the major version and architecture match
between nodes, it should be safe to use binary copy.

-Dave