I wrote:
> As this example shows, we really ought to validate the compression
> argument on sight in order to get sensible error messages. The
> trouble is that for server-side compression the code wants to just
> pass the string through to the server and not form its own opinion
> as to whether it's a known algorithm.
> Perhaps it would help if we simply rejected strings beginning
> with a dash? I haven't tested, but roughly along the lines of
Taking a closer look, many of the other switches-requiring-an-argument
also just absorb "optarg" without checking its value till much later,
so I'm not sure how far we could move the needle by special-casing
--compress.
regards, tom lane