Re: Debunking a myth: Knuth-Plass as a silver bullet for widow/orphan management

"G. Branden Robinson" <[email protected]> Wed, 12 Aug 2026 05:34:55 -0500

Newsgroups	gmane.comp.printing.groff.general
Message-ID	<20260812103455.mxcaaf5bcnniyjsr@illithid>

At 2026-08-11T18:17:53+0200, Stefan Möding wrote:
> Tadziu Hoffmann <[email protected]> writes:
> > TeX's formatting model is based on optimization, finding the
> > "least bad" instance of a line-broken paragraph.  How "bad"
> > orphans ("clubs" in TeX) and widows are considered to be
> > is user-specifiable.  If you don't set a sufficiently high
> > penalty for these, then other criteria will dominate, and
> > TeX will in fact produce output with stranded lines.
> 
> The TeXFAQ at https://texfaq.org/FAQ-widows says “These penalties are
> usually set to the moderate value of 150; this offers mild
> discouragement of bad breaks.”
> 
> Since the TeXbook also mentions 150 as value for these two penalties,
> this obviously goes back to D. Knuth himself.  I’m pretty sure that he
> spent some time thinking about the default value.

Yes, but Tadziu _also_ said:

"The whole business of typesetting is finding the right compromises
for a given job."

Different circumstances and contexts dictate different compromises to
the artist's eye.

The quality of Knuth's output in typesetting his own work is tough to
criticize.  I can say only that when I read him, I don't notice the
typography, which may be the highest compliment I can offer it.  He gets
me thinking about math and/or engaging with his approachable authorial
voice; I don't perceive the scaffold upon which he arranges his words.

I _did_ notice subpar typography in the C++ paper I cited in the
original post to this thread.  That's no shade on the authors--these
types of papers are intended to be ephemeral and eventually will be of
interest mainly to archivists and historians of that programming
language.

My point, apart from pointing out the hazards of ignorant fanboy culture
that I see at various web sites, is that both TeX and *roff offer the
author the tools they require to achieve good typography.  Lacking an
implementation of the Knuth-PLass breaking algorithm doesn't mean a
*roff can't do that.  It's not a killer feature.  Heirloom Doctools
troff and neatroff both wasted no time in implementing it.  I don't know
that the latter had or has any ambition to supplant groff; I harbor a
suspicion that the former did.  If so, that has yet to eventuate.

Regards,
Branden

signature.asc (application/pgp-signature, 833 B)