Re: [Groff] RE: Small bug in groff 1.19.2 footnote number contro

[Ralph Corderoy]

Hi,

Gunnar Ritter wrote:
> Keith Marshall <address@hidden> wrote:
> > > It's a genuine troff mm-ism.  E.g. we found out about it from the
> > > book by Narain Gehani (of AT&T) "Document Formatting and
> > > Typesetting on the UNIX System", ISBN 0 -9615336-0-9 (highly
> > > recommended, BTW).
> >
> > This may establish `prior art', but it doesn't constitute official
> > documentation; there is no onus on groff, to replicate this
> > undocumented feature of another troff implementation, and I agree
> > with Mike -- it shouldn't do so.

The official documentation isn't a complete spec. by which a
re-implementation could be created that would work well with existing
documents.  The source was available to those using the macros and I
can't be the only one that used to grovel through the "no spaces, no
comments" versions to understand how to get the desired behaviour, i.e.
what registers were used for what.

I had no contact with other trofflodytes back then.  Those in
institutions must have shared their knowledge of undocumented behaviour,
even had "tip sheets" of such things.  If Gehani documents it then it's
well-known and "official" AFAICS.

> But on the other hand, the original -mm documentation (which is the
> only text that could count as authoritative this way), "Memorandum
> Macros" by Smith and Mashey, actually encourages using undocumented
> registers:
> 
> In section 14.1.2, it states that registers of the form ":x" are
> "mostly internal, rarely accessible, usually dedicated".
> 
> But then, there is an Appendix A, "User-defined list structures",
> which contains the advice: "To understand it [i.e. the new list
> macro], you should know that the number register :g is used by the MM
> list macros to determine the current list level". But :g had never
> been mentioned ("documented") in the preceding text.

It's great to take a decades-old document and find it still formats
adequately in modern-day groff.  To loose that in trivial cases, like
aliasing :p, or implementing the :C reset, seems such a shame.

I vote yes, maintain compatibility where straightforward.

Cheers,


Ralph.