On Mon, Jun 14, 2010 at 1:18 PM, Michael Hanselmann <[email protected]> wrote: > Am 14. Juni 2010 13:08 schrieb Guido Trotter <[email protected]>: >> On Mon, Jun 14, 2010 at 12:28 PM, Michael Hanselmann <[email protected]> >> wrote: >>> The hostname and port received from the remote cluster should >>> be validated, just in case. >> >> This is not the same patch, did something happen? > > Yes, something must've gone wrong, sorry. Will resend the correct patch. > >> Also, small question: if the remote cluster passes a service name, how >> can we guarantee it exists on the local cluster? >> Should we look it up in the services db? Or should we enforce a >> numeric port to be passed? > > They get what they ask for: breakage. Don't try to overdesign this > function. It's simply validation and so far the port is always taken > from the import-export daemon (which takes it from socat) and sent to > the remote cluster. >
That's the point, so it's always an integer, now, correct? Wouldn't requiring an integer inside the port numbers range make things simpler, and avoid overdesign even more? Is there any case where a string is needed, currently, or where you foresee a string being used in the very near future? Guido
