Not to engage in whataboutism,
but to engage in whataboutism,
what about the groff language default of 1 for the .hy request?

Today's episode: boring into the Boringest Conspiracy Ever

comment #4:

> it doesn't matter if Russell Harper tells us that 99% of all literature
> published since 2000 by the University of Chicago Press follows the single
> intersentence-space rule if a representative sample can be taken of those
> publications and a statistically significant different proportion turns up.

100% agree.

> And that sort of empirical measurement is what I don't see in these arguments.

True.  We're taking a lot of authorities at their word because we don't have an independent commission, perhaps a branch of the the Department of the Interior, whose mission includes taking typographical measurements across publishers and over time.  We lack chart-filled reports with titles like Practices In Kerning: 1970-2000.

I'm gathering that your response to this dearth of data is to throw our hands in the air, declare "All is unknowable!," and either wait for typography's Robert Mueller to issue his report, or dive into the field and gather the data ourselves.

My proposal is simpler: let's give the authorities the benefit of the doubt until data emerges to call their authority into question.

If someone wants to claim there is a discrepancy between style-manual edicts and practices by publishers claiming to follow such style manuals, the burden of proof pretty much has to be on the person making that claim.

And I don't think accepting the authorities' statements at face value is going that far out on a limb.  Forget careful measurements for a moment: just look at some modern examples of professionally typeset works (i.e., not the garbage that Word or the web produces).  You'll be hard-pressed to find examples with discernible wider sentence spacing.  Not even in The New Yorker, known for its idiösyncratic style guide.

Thus, even if there is some additional sentence spacing that's not easily visible to the naked eye, it's at levels on par with ".ss 12 3" or ".ss 12 4".  Groff's default, ".ss 12 12", is grossly out of step with that.

So we have authorities telling us that word spaces and sentence spaces should be the same, and we have a corpus of works published in the past 75 years, where, by gum, they all pretty much look the same.  Ockham's razor strongly suggests that using normal word spaces between sentences is today's industry standard, and has been for decades.

While we're waiting for concrete data, groff has to have some default for sentence spacing.  I submit that the most sensible one is the one that pretty much every modern authority agrees on.  If and when data surfaces that shows these authorities to be widely ignored, the question should absolutely be revisited.

comment #3:

> I think this passage warrants an additional response.

Sure.  I don't see much I can fruitfully respond to in comment 2 that I can't sweep up with this one--I like to get concrete, and I find the virtue-preachings of authority figures less persuasive than empirical measurement.

> comment #1:
> > I think we should liaise with the TeX community before proceeding along such an iconoclastic path.
>
> But it is groff's and TeX's defaults that are the outliers on this.  (I didn't know till a quick google just now that TeX also defaults to wider sentence spacing.)

TeX and *roff (historically, a mix--today, mostly TeX, I think) enjoy a status in many fields of technical writing whereas they are almost completely unused elsewhere.  This boundary is meaningful: it represents community, shared practice, some degree of shared values, and moreover this boundary is constitutive of a discernible and justifiable set of independent idioms.

To be less abstract, there are several different citation styles in English professional writing: legal, MLA, APA, Chicago/Turabian, and so on.  By the same argument you are making here, you could decree that groff should support by default the single most popular of these.  Perhaps even if there were only a plurality winner, meaning a majority of users would be disserved?

Community matters.

> As an academic exercise, it would be interesting to know TeX's reasoning.  As a practical matter, it's irrelevant.  The convention used by the rest of the world is clear; groff's and TeX's defaults do not align with it.

It doesn't align with, arguendo, the vast majority of authorities writing style guides, in English, outside the hard sciences.

That does not mean that how digital type actually gets set, even in specimens that these authorities would declare normative, adheres to those prescriptions.

This subject is so full of puffery and grandstanding (to which I, admittedly, may be contributing my own vituperation) that I am suspicious of the claims being made for the prevalence of single inter-sentence spacing.  Here's why.

Again, arguendo, I'm willing to stipulate to an overwhelming prevalence of identical spacing between words and sentences for the extremely common case of filled-but-not-adjusted (i.e., ragged-right) text composed in Times New Roman using Microsoft Word (or LibreOffice Writer) and then composited into a box on a page (often directly to PDF by the same user) from there.

Such documents may comprise 95% or better of all "typeset" documents ever produced, such as informal memoranda, though much of that has shifted to email over the past two decades--as a matter of fact, let me interrupt my own argument here.

I submit that a large part of what drove the uptake of paperless memorandizing in professional contexts was the advent of "rich text" email.  You couldn't get people who weren't, at some level, computer nerds to adopt email until people could put real boldface, italics, and underlining into the text (often, all three at once) so that the excitable (or, less happily, authoritarian) could place appropriate emphasis on their mandates and prohibitions.  (To level up your migraine to a cluster, change the font to Comic Sans and add greengrocer's apostrophes.)

Among the many possible candidates, what technologies were used to achieve that transition?  "Rich Text", for a while stored and transmitted as "RTF", which as I recall was a feature-limited version of a format originated by Microsoft Works (a hobbled version of MS Office at a lower price point), and, shortly thereafter and tellingly, HTML.

HTML is notorious for its (non-)handling of whitespace.  Of course you were going to get only one intersentence space in HTML. Type one, type 50, mix in some tabs, one space was what you were going to get.

Rich Text and HTML are typography, but of a particular, limited, not to say debased, variety.

I expect you to tell me that even if I'm precisely correct, my argument is irrelevant.  Don't worry, I haven't forgotten and will return to it.

> > That community and ours are the only ones I trust to produce well-reasoned opinions in this field.
>
> Perhaps that's the core of our disagreement: I don't think it matters how well reasoned anyone's defense of wider sentence spacing is.  You could give me 100 reasons why it's better, and even if I agree with 99 of them, it doesn't change today's industry standard.

It doesn't change what people are saying, certainly.  But are we examining professionally-set documents and taking their measurements?  And by "we", I mean real researchers.  Societally, are we just taking Russell Harper's word for it?  Sure, he's an authority, but the way authorities achieve their status is by building a peer-reviewed corpus of findings.  An authority whose claims you cannot interrogate (albeit with appropriate training or the aid of someone who has it), is a false one.

The non-adjustability of groff's own intersentence spacing was, apparently, little remarked-upon until you came to the issue and it remains hard to observe, especially in proportionally-spaced fonts, until one contrives rendering parameters to reveal it.

Bluntly, I don't trust most of the authorities you cite to experiment the way we do.  Before you say that doesn't matter, either, I would remind you that majoritarian arguments about what is the case have to be independently derived if they come from sources that confuse their wishes with reality.

For example, it doesn't matter if Russell Harper tells us that 99% of all literature published since 2000 by the University of Chicago Press follows the single intersentence-space rule if a representative sample can be taken of those publications and a statistically significant different proportion turns up.

And that sort of empirical measurement is what I don't see in these arguments.  Possibly, when it comes down to the nuts and bolts of page composition for publication, there is a community of experts who apply their own practices regardless of what the old man in the glass-walled office upstairs says.

I admit that that is pure speculation.  But somebody has to be measuring these things, don't they?

And if they're not, why should we take these exhortations as anything more than aspirational?

> I certainly can't impugn TeX's typography, regardless what I think of its syntax.   But trusting only their opinion, while dismissing those of typographic experts like Bringhurst, the designers of commercial typesetting packages, and the publishing houses who use them, feels a bit like choosing your allies based on the answer you already know you want.

Where are the designers of the commercial typesetting packages on the record?  Where are the configurable parameters for these packages by major publishing houses documented?

I'd like to see these for curiosity's sake, at the very least.

> (As a possibly interesting side note, if I understand http://tex.stackexchange.com/a/4726 correctly, TeX's default is a sentence space 33% wider than a word space, whereas groff's default is one 100% wider.  So groff's double-wide space is already out of step with TeX's much subtler widening.)

So in groff, for a typesetter/troff device, we could express that as:

.ss 12 4

The above won't work (by my lights) for typewriter/nroff devices because there are is no fractional spacing; groff will round that 4 down to a zero.

Let me return to the majoritarian argument as promised.

The majority of groff documents are not written in the raw language.  I myself never wrote a single such one until I joined the development team (and now I do it all the time to test things).

They are written in macro languages by package which take a variety of approaches to intermix of their own lexicon with that of the underlying roff engine.  So I suggest that a good place to set defaults, as with hyphenation and adjustment modes, is the macro package.  Long ago I wondered why our macro packages (and TeX's) didn't have modules for APA, MLA, and so forth.  (The answer appears to be a combination of long-toothed packages that have seen little change for decades, and the relative dearth of use of either *roff or TeX outside the hard sciences.)

If you want to get people's groff documents conforming to a set of common practices, then a good way is to provide a macro package that encapsulates them.  I don't propose that you (or we) alter or rewrite ms or mm.  We could start small.  As documented in the Texinfo manual, there are four fundamental things groff does with text:

• filling
• hyphenation
• breaking
• adjusting

The .ss request directly affects the very first of these.

So a first cut of "chicago.tmac" might look like this:

.fi
.ss 12 0
.hy 6 \" no idea--what does the CMoS say?
.ad l

(My copy of CMoS, a book I otherwise highly esteem, unfortunately did not make it with me in an intercontinental move.)

We then tell newcomers to groff who seek to produce workaday documents to pass "-mchicago" as an option.

We certainly wouldn't tell them not to use a macro package at all; that's been a bad idea since the 1970s.

groff the engine has long catered to a specialized audience of technical writers in particular fields.  It may very well remain there without a GUI front end for document composition.  (I find LyX impressive, personally, and have used it for real work.)  But it is designed to cater to the broadest expressible class of typographical challenges, including non-textual drawings and non-English text.

I think I find your majoritarian argument unpersuasive because it poorly fits both the majority of uses to which groff is actually put at present, and the, shall we say, aspirational majority of typographical ends to which it could be put, which includes such things as the annotation of architectural schematics in Mandarin Chinese.

I have assigned this ticket to myself because I am interested and I am, indeed, seeking more information as the status indicates.  If the matter comes down to "will the groff default for the second argument to .ss change in src/roff/troff/input.cpp (or wherever)?" change, it is my intention to unassign myself from the ticket rather than bottling it up as a sort of pocket veto.

Right now, I oppose such a change, and would -1 it on the mailing list (however we interpret such a thing in our community), but I do not wish to be anything more or less than a non-volunteer for such work.

Thank you for the discussion, by which I certainly do not mean to draw it to a close, but rather to reassure you than I am not personally piqued and continue to appreciate your collegial work on groff.

comment #1:

> While I think you're right to characterize the growing consensus among professional writers who use computerized typesetting systems but don't develop them, I don't think groff should be bound by that consensus.

That's not how I would characterize this consensus.  I'm not looking at who is making the decisions, but at the decisions themselves.

As of 2020, with very little exception, major publishers use the same spacing between words and sentences.  We can't peer into those publishing houses and tell if those decisions were made by professional writers, software developers, or janitors.  All we can do is look at the results, which tell us that major publishers today follow this practice.  This makes it the current industry standard.

You may hate it.  I'm no fan of it myself.  But no less a typographical authority than Robert Bringhurst endorses it in The Elements of Typographic Style.  His opinion carries a weight that yours and mine don't.  Pretty much every modern style manual agrees: there's an entire Wikipedia article documenting this consensus.  (And given how many details these various style guides diverge on, the unanimity on this issue is remarkable.)

It's hard to dismiss every major publisher and style guide as guilty of typographic ineptitude.  Ultimately, it doesn't matter whether the current standard arose from ignorance, or as a cost-saving measure, or as the result of a long-ago meeting of the Secret Cabal of Typesetters whose minutes we are not privy to.  Even if you claim all these publishers and guides are "wrong," collectively they define the industry standard.  "Wrong" is the new right.

Groff's default sentence spacing is out of step with the industry standard, and that's a bug regardless how much you or I may wish the industry standard were something other than what it is.

As mentioned in passing in bug #58450, typographic convention of the last 70-100 years has been to use the same amount of space between sentences as between words.

groff allows the user to specify any amount, including none, of additional inter-sentence space to add.  This flexibility is good.

However, by default, groff should follow modern typographic convention.  This would require the default value of sentence_space_size (register \n[.sss]) be 0.  Instead, currently:

• If a document never calls the .ss request, groff defaults to a value of 12 for word_space_size and 12 for sentence_space_size.
• If a document calls the .ss request but with only one argument (word_space_size), groff sets sentence_space_size to be equal to word_space_size.

In both these cases, the default sentence_space_size should be 0.