Re: transaction idle timeout

=?ISO-8859-1?Q?Szima_G=E1bor?= <sygma(at)tesla(dot)hu> writes:
> Here is a patch for backend to allow transaction idle timeout.

I don't think this is a good idea in the first place. But even if it
were, this patch is a mess. secure_read() has no business doing the
things you have made it do, either messing with the PS display or
directly calling AbortCurrentTransaction; the patch doesn't work for SSL
connections (and in fact probably actively breaks them); it breaks
handling of all I/O errors other than timeout; it significantly
increases the load imposed by an idle connection (since your
implementation causes a waiting backend to wake up once a second, even
if the feature isn't in use); and there's no documentation.

regards, tom lane