[Groff] Progress report on the portability audit -- and what to do about URLs?

[Eric S. Raymond]

I am about three-quarters of the way through cleaning up the groff
manual pages to be viewer-portable.  The new versions will also
be simpler, shorter, and easier to read than the old.  I've done 
these so far:

chem.man
pdfroff.man
ditroff.man
groff_diff.man
groff_out.man
roff.man
groff.man
grog.man
nroff.man
troff.man
an-old.tmac
groff_me.man
groff_trace.man

I don't anticipate serious problems with the rest of the work, which
will probably take about two more days.  Mostly it consists of ripping
out a lot of over-elaborate macrology and coding simple replacements.
This has not proven difficult.

There is one major exception. Many groff pages use the www.tmac macros
.URL and .MTO.  This opens up a significant can of worms because the
www.tmac code is unsalvageable, full of constructions that cannot be
made viewer-portable.  Thus, any page using these will either break
some viewers entirely or at best display in a somewhat mangled form.

I would have caught this sooner, but doclifter just ignores the
www.tmac include and interprets .URL and MTO itself.  Therefore it
doesn't have a problem, and if I were only trying to solve doclifter's 
issues I could declare victory and go home.

We have several possible paths to go from here:

(1) Rewrite www.tmac in portable code (no groff extensions).  This
   could work because most third-party viewers do follow .so directives.

(2) Add portable implementations of .URL and .MTO to an-old.tmac

(3) Restore the .UR and .UE macros that used to exist in
   at least in the Linux version of the man macros.  I can't find the
   documentation for these any more, but I remember reading it and
   support for it is in doclifter; what happened to this feature?

(4) Rip out the MTO and URL macros entirely, and trust Web-aware viewers
   to pick out URLs and mail addresses by pattern recognition.  Some,
   including doclifter, already do this.

I included (1) only for the sake of completeness; I don't think it's 
a realistic alternative.  I have no strong preference among the other three.

I need a quick decision about this, because it will affect the set of cleanup
patches I turn in.
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>