Re: BUG #6300: duplicate key value violates unique constraint

On tor, 2011-11-24 at 11:14 -0500, Robert Haas wrote:
> So I would propose to steer clear of the word "internal", because the
> really scary errors typically are not internal to PostgreSQL at all.
> What I think we want to distinguish between is things that are
> PEBKAC/GIGO, and everything else.  In other words, if a particular
> error message can be caused by typing something stupid, unexpected,
> erroneous, or whatever into psql, it's just an error.  But if no
> input, however misguided, should ever cause that symptom, then it's, I
> don't know what the terminology should be, say, a "severe error".

The current error levels are designed entirely in terms of the client
session behavior, that is,

warning -- things continue
error -- abort transaction
fatal -- abort session
panic -- abort everything

(more or less).  For a client issuing statements and reading responses,
these levels make perfect sense.

What we need is a labeling system in terms of server behavior, which is
completely separate from these client levels.  In principle, every
log-issuing statement (that is, ereport) should specify a client and a
server severity level.