As it happens, I was hacking a bit on Heilhveiti (a renamed Vollkorn) a short while ago. Its design height is 1000 units (750A+250D), so one can take that as its em height as well. Its space (#32) is 200 units, making it one-fifth of an em.

http://www.microsoft.com/typography/developers/fdsspec/overview.aspx is a Microsoft typography article about recommended settings and shapes when making Latin-based type. There is also mention of French typography.

In short, it seems to be a matter of preference.

Back in Ye Olden Days, before the democratizing of typesetting, dedicated phototypesetting systems generally made numerals en-space width and punctuation thin-space width. Since WYSIWYG was a far-away dream, this arrangement made setting tabular material a little easier and far more predictable.

So, fast-forward to today: use a period to indicate a thin space and set its fill to 0% black/paper/whatever…

Té Rowan, thank you for the great page about spaces. This explained everything and also how come I could type a 1/4th em and get a bigger space than a normal space in the font I was using earlier today! Looks like this theory I have needs revision!

Oldnick, I like your tip although I'd be afraid they would forget to change the color at some point and make a mistake. I like that it will be adapted consistantly with whatever type used and it's relation with history. I'll mention it as well.

I still don't know how kerning units are calculated... anyone? :-)

If inDesign lets you and the font has them, you could use some of the Unicode spaces (thin space, punctuation space) for this purpose. This, by the way, is why I was mucking about with Vollkorn/Heilhveiti in the first place – to toss in some nice-to-have spaces and dashes.

Re kerning, if a font has it, it's done in design units. A comment on an inDesign help page indicates that iD sets kerning in 1/1000em increments. So, 1/8em oughta be 125 units. /me scratches 's head.

A thin space has never been any one thing. If you ask an old Linotype (linecaster) operator, he'll tell you a "thin space" is an unexpanded spaceband. The linecasters weren't on an em system; if thin spacebands were used, an unexpanded spaceband was about 2 points. IIRC, Monotype was a 5-to-em space. Some took it as 6-to-em. The Chicago Manual at one time specified a thin as a 4-to-em, but no good comp I knew ever used that. Oddly enough, in those days Chicago used someone not at Chicago for there typesetting/printing definitions, and that person (a designer) happened to be at odds with the world. See Glossary of Typesetting Terms.

* * *

I believe the way InDesign works is if the space characters aren't in the font (most fonts lack them), InDesign uses a formula. It is documented. If the spaces are in the font, InDesign will use them. Unicode now defines the values, though of course, you can use what you want. I usually use about .055 em for a hair space, and .160 for a thin (close to 6-to-em). The 3-to-em, 4-to-em and I believe 5-to-em are also provided for by Unicode, in addition to "thin" and "hair"

BTW, CS4 for windows had a nasty bug. If (1) you did have the space characters in the font, and (2) used 9 of them on a single line -- say, ens, as happens on a cip page, and (3) had "view invisible characters" (close to that, anyway,) "on," the program would crash every time. Miguel Sousa at Adobe checked it out, and reported that bug isn't in CS5, and wasn't in CS3. Life is strange.

Related discussion: http://typophile.com/node/68985

Comment specifically about InDesign’s formulae: http://typophile.com/node/68985#comment-404470

Thanks Kent. I had forgotten about that thread. It is interesting, the internet threads give a lot more information than a vetted, published source like the Glossary. Most of it correct, and with the bonus value of Nick's sarcasm.

I still don't know how kerning units are calculated... anyone? :-)

Visually is the most effective—and time-consuming—way. Perhaps DTL Kernmaster 2012 will break the mold, but most of the algorithmic kerning programs or functions now available tend to make a number of bad decisions, which need to be undone manually.

Another flashback to Ye Olden Days of Dedicated Typesetting Machines: the fonts available did not contain any kerning data, as a rule. Kerning was done after all of the keystrokes had been input by a simple search-and-replace function, on the assumption that the limited number of search pairs would cover most instances. In general, this was true, but one always had the option of going back in and modifying the source file if the output shouwed any glaring errors and/or omissions.

Well, after some testing the kernings units do correspond to 1/1000th of an em. The theory is just wrong! :-/ Gotta love being a new teacher and challenging what's taught already. I'm proud and feel out of place at the same time, haha...

I like the tidbit about the Chicago Style Manual... maybe this is where our 1/4th of an em comes from.

As usual, all type answers can be found on Typophile and so much much more. Thank you!

This was too easy... :-)

Just to confuse me even more...
In the english help file of InDesign :
Quarter Space...One‑fourth the width of an em space.
and
Hair Space...One‑twenty‑fourth the width of an em space. So far so good...

and its equivalents in the French file :
Quart d’espace...Un quart de la largeur d’une espace cadratin.
and
Espace 1/4 cadratin...Un vingt-quatrième de la largeur d’une espace cadratin.

It's word for word the same text except for the name Hair Space is changed to Space 1/4 of an em! Argh! Maybe they just forgot the 2 in 1/24.

I must confess that I fail to see the relevance of those values for the spacing in French punctuation. Indeed all those values are defined in terms of the em size; however, the eye can't judge the spacing before a semicolon or an interrogation point in terms of the em size; all it can do is compare it to close by inter word spacing; with LaTeX, the inter word spacing is easily accessible as \fontdimen2 and I feel no guilt in redefining the French spacing used by the babel package in terms of it (and other easily accessible parameters), especially with fonts for which 0.16667em (the LaTeX thin space) is almost equal to the inter word spacing.

Here are a few examples that show more clearly the problem of defining thin spaces (or at least punctuation spacing) in terms of the em size. In all the fonts I looked at, the em size is 1000. In Arno Pro Regular, the width of the space character is 190. If you define the thin space as the fifth of the em size, you get 1000/5 = 200, and a thin space that is then larger than the space. With 1000/6 you get 167. With 1000/8 you get 125, which is 66% of 190 and may seem acceptable to some.

However, if I run a script to get the width of the space character in other fonts I got from CS3, I get still smaller values for the space width:

ArnoPro-Display.otf 158
ArnoPro-BoldDisplay.otf 150
MyriadPro-CondIt.otf 147
NuevaStd-Cond.otf 129
NuevaStd-CondItalic.otf 104

With Nueva Condensed, the space width is approximately equal to 1000/8, not to mention the Condensed italic.

@emspace - Yeah, that's a right effin big blunder in the French file.

And to really set your eyes swirling, the French Wikipedia page on spaces says that a thin space differs between English and French typography (em/4 v em/5).

I think... I think it's time you start your own school on proper typography.

Aside: The English Wikipedia page on sentence spacing makes some interesting comments.

Here are a few examples that show more clearly the problem of defining thin spaces (or at least punctuation spacing) in terms of the em size. In all the fonts I looked at, the em size is 1000. In Arno Pro Regular, the width of the space character is 190 . . .

Which is too bloody tight. The compositor can change the value of a nominal space in InDesign by changing the percentage from 100% to another value in the justification menu. Or, since it is an Adobe font, go into the font and change the value there. In an ideal world, this wouldn't be a licensing issue -- the last time I used TeX, the spacebands used were not from a font character.

* * *

Elsewhere on Typophile, Kent Lew has given his technique for determining the spacing values to use in InDesign with a particular font -- I believe he started by setting a page worth of text ragged right, then print it out, evaluate, and adjust. Of course, the values picked are subjective.

But you're right, to have the word space subject to the whim of the type designer but have the other spaces "fixed" by convention is an oddity. Again, if you go into the font, you can change the vales of "thin" and "hair." (Well, you *could* change them all, but you'd fly in the face of defined fixed spaces.) Anther oddity of fonts, perhaps a legacy from Type 1 fonts where fixed spaces were rare, is giving the nobreak space character a value of 3-to-em, or even 2-to-em.

* * *

For teaching, the best advice is to inform students that all spaces in typeset material interact. For printed texts, that would include margins. Digital editions lack a true page, so page/margin spacing tends to disappear. But the relationship of character spacing, word spacing, and line spacing remains. The "proper" space around punctuation is in large part set by these relationships.

For those that read French, here is a relevant comment from Jean-François Porchez: http://www.typographe.com/forum/viewtopic.php?id=715

Charles: the last time I used TeX, the spacebands used were not from a font character.

If you simply use XeLaTeX with Arno Pro and the polyglossia package, here is what you can get in a two column text (after some automatic shrinking of the interword spacing).

ps: In the input file, there is no spacing before the semicolon.

Michel, that looks like crap. The interword space is too tight, the space around the punctuation too loose. For generalities, IMHO, Porchez has it right.

(The TeX I used was plain TeX with our own macros, including a page layout set.)

Michel, thank you so much! This was the last piece missing for the puzzle. I had actually written to Mr. Porchez just the day before you posted this so it looks like I was right in thinking he could answer me :-) I got his reply this morning and he directed me to this article he wrote : http://www.porchez.com/article/272/Lesespacesparentspauvres

It's in French but overall explains how after Didot, interword spaces used were 1/3em and larger. At the time, the thin space was generally 1/4em. Afterwards, when the interword spaces shrunk, some people didn't change the thin space (duh!!)

I'm almost done rewriting all the theory we had on the topic. I was supposed to give it tonight but I still have to experiment on the best way to fake a thin space in Illustrator so I'll give it next week.

Émilie

Emilie

If you search for ponctuation the Banque de dépannage linguistique of the Office québécois de la langue française, and then select their text about http://spacing before and after punctuation signs, you get in the first paragraph their definition of espace fine namely une espace insécable réduite (a reduced non breaking space). For them spacing may be either normal or insécable and by the context, they are clearly referring to the interword spacing.

Michel

I do use the Banque de dépannage linguistique although I wasn't aware of that specific page, thank you! The problem is because of the school setting they require to use a book. Before, we used the Ramat (ouch!) and now there is a Manuel de Typographie published by 2-3 teachers I think. Overall it's a very nice manual and I don't want to complain too much because it's already difficult to find French text books in graphic design. But the definition in the manual for a thin space is that it's equal to a quarter em. I had the library order the Manuel de composition typographie by Muriel Paris and maybe I will suggest it to the department but I don't think it covers the history and such.

@oldnick : I've been experimenting with thin spaces in Illustrator and your trick with the period works well, in fact it's the only way to type it so it will be non-breaking that I am aware of. But I'm not always getting the same space (tried it in Minion, Myriad, Archer and the period space in Archer is thinner than the other fonts) and still trying to figure out how to have a flawless 1/8 non-breaking em in Illustrator.

/just realized there a "no break" option in character palette of Illustrator!

Nevermind then! So I guess I found two decents ways for creating a non-breaking thin space. Kerning the letter and punctuation sign by +125 or inserting a space at 50% horizontal scale and applying no break with the previous word. Then it will still be selectable. It seems to work fine on all the fonts I've tried so far.

Finding the best looking amount of space is not the same as trying to translate yesteryears thinspace to what InD can understand. Different fonts have different kerning. Some may have built in space to accommodate the French preference either in default kerning, with contextual kerning, or even using language specific contexts. It is always best to look at the actual font you are using as it comes from the foundry before deciding if you want to augment the spacing. After testing in the language and text you are dealing with, you can see what needs to be done. You might also make a note to yourself about which fonts need what adjustments and make a style sheet for later use.

The colophon of my 1975 edition of the Lexique des règles typographiques en usage à l'imprimerie nationale, that has been the reference for many years, says it was printed in monotype Baskerville by Imprimerie nationale. Presuming they applied their own rules, I looked at a few of their semicolons and started doubting their thin space was really as much a fourth of an em. Now, the top is a picture of the end of a line (page 38), so that no space has been squeezed or expanded for justification. At the bottom is the Monotype Baskerville I have on my mac. The point size I used is 10. To get the same spacing as the original, I had to add a kern of 150 before the blanks; and to get the semicolon where it is, I had to add a kern of 80. Now you tell me the size of the space added before the semicolon; I don't know what units inDesign is using.

Michel

I forgot to add that the em size of the font is 2048 and the width of the space character is 512.

From what I have seen here and elsewhere, Indesign uses milli-ems. I'd guess from your experiment that their 'thin' space is ca. twelve to the em and the word spacing is two-fifths of an em.

Sounds like me and my pocket calculator are both gone cuckoo for cocoa puffs.

If you are using InDesign, you will notice that, for appearances only, it acts as if all fonts are on basis 1000 when it comes to manual kerning and displayed values. Apparently, InD calculates (with rounding) what a 2048 font would be if it were rounded down to one thousand units. This is what is displayed in the InD kern value. I assume it does not really change the actual value but just shows you a conversion factor value.

If those units are on a basis 1000, then 1000/12 is 83 which is about the kern after the r preceding the semicolon; that is indeed twelve to an am, as Reynir got. That may be an indication that four to an em was not the intended meaning for the espace fine to be put before some punctuations marks. To better see what their actual usage is (or was), I think the same experiment should be carried on some of their most beautifully edited books.

Here is from an edition of Pantagruel, Paris : Imprimerie nationale, 1990, c1989. According to the colophon, the book was hand composed using a Garamont character also known as "Romain de l'université" (Le texte de Pantagruel a été composé à la main à l'aide des caractères Garamont également connus sous le nom de « Romain de l'université ». L'ouvrage a été réalisé par l'atelier du livre). I am not trained to guess the size of the spacing before the semicolon.

I have here a large book of miscellaneous occasion poetry printed in 1945. Looks to me like the text is en-spaced. Were it used, the normal thin space would indeed be thin in that book.

Just for fun, here is a small sample from this book (a part of the author's redo of Longfellow's "The Arrow and the Song") translated into English by myself:

"I shot an arrow into the air,
and it flew away to the-Devil-knows-where.
A raven sitting on a weathervane
thought it was an aeroplane."

@Michel: …so that no space has been squeezed or expanded for justification…

Not so.
The word spaces of final lines were often made wider to balance out with the average word space of the entire paragraph/column, which is wider than a type space, due to justification.

@ Nick.
That makes sense.
On the other hand, I notice that the comma after Signis is italic but all the semicolons are upright, even after an italic. Would the spacing between the word and the semicolon be the reason?

ps. Or it might be the opposite. The spacing might be there to justify an upright interrogation mark or an upright exclamation mark after some citation in italics?

By the way, here is a similar treatment of semicolons in a dictionary published by the Oxford University Press, (c) 1963.