Our algorithm for anonymous step names doesn't work

Our algorithm for computing step names for anonymous steps doesn't
work. If you import two anonymous pipelines, they both have the same
anonymous name "!1". No doubt there are other ways to run afoul of the
fact that the naming is only unique in the context of a single
document.

We could say this doesn't matter, since the anonymous names can't be
used in any pipeline, but that seems like cheating. Also, it means
that the step name used in an error message might be ambiguous.

I suspect the cost effective answer is to punt and make the names of
anonymous steps implementation-defined.

                                        Be seeing you,
                                          norm

-- 
Norman Walsh <ndw@nwalsh.com> | A rusty nail placed near a faithful
http://nwalsh.com/            | compass, will sway it from the truth,
                              | and wreck the argosy.--Sir Walter Scott