My point of view has always been that you shouldn't change a glyph. A typeface is designed by someone to work a certain way. Usually, if you impose your own new parameters on their design, like condensing it or expanding it, it won't work as well anymore.

Several font families have taken this into account, and designed fonts with multiple widths (compressed, condensed, regular, wide, expanded, etc.). If I needed to alter a font's width, my first inclination would be to see if there was already a similar font with the width I needed.

In the photo-typesetting era, people first realized that they could mutate letterforms (Matthew Carter called this "mutable type"). The results were almost universally terrible. Gerard Unger has a popular lecture in which he describes how a collegue of his was required to mutate type weight, and the results were so bad that he almost cried.

Long ago, when Matthew Carter was working for Mergenthaler Linotype, he designed a Univers-like family called Video for photo composition. This was a typeface designed to take abuse. Shortly after it was released, photosetting machines evolved to the point were it became easier and less costly to switch between different fonts, and Video was withdrawn.

Under "optical scaling" I think of something completely different. Some type families now, Minion for instance, have different weights that were designed for use in different point sizes. You use the normal Minion for text, and the Titling weight for display. In theory, you could design 20 different optical sizes instead of just two, which would make digital typesetting like cold metal typesetting used to be.

Without trying to sound too preachy, I would say that using optical sizes is generally viewed by graphic designers and typesetters as a "good" thing, while mutating typefaces (i.e., artificially altering their width or weight) is generally viewed as a "bad" thing.

The idea of intelligently scaling the width of glyphs to aid line justification was part of the Hz system developed by Hermann Zapf with URW, and later purchased by Adobe, who have implemented some parts of it in InDesign. It is discussed in Bringhurst's Elements of Typographic Style. The kind of horizontal glyph adjustment suggested by Zapf is different from typical, universal scaling: the idea is to subtly alter the proportions of letters without affecting stem weight in such a way that the overall colour of the text is altered.

Does that mean that InDesign is scaling my glyphs when I justify text? I thought that it was just changing my word spacing.

I don't think InDesign implemented that part of Hz.

I think there is an option somewhere in indd to change width of letters themselves on spacing prefs. More basic than Hz, perhaps, don't know, I have only a printed sample of the original Hz stuff set in one of H Zapf typeface.

Dan you have to specify if you want ID to scale type. I personally don't like changing the glyphs from how they were drawn. I can always change my layout, I'm not that unflexible.

Greetings #42,

See also this message on the FontLab forum.

> I strongly prefer the virtually-unnoticeable
> stretching over having too-wide spacing
> between words.

Agreed, although Design being the balance of compromises, ideally you really want to bend each factor (letterforms, wordspacing, letterspacing) so the whole is balanced.

hhp

There is a lot of precedent for alternate letterforms to aid justification. One of the best and most subtle examples would be the narrow and wider 'r' alternates in Robert Blumenthal's 1931 Spiral. See plate 21 in Pankow's American Proprietary Typefaces for a striking example; the difference is practically invisible, but it's also easy to see how it works to improve spacing.

Another seminal reference is the Thermo type family of ATF, featured in their 1934 catalog. In this font, there were three separate courses of capitals and numerals, carefully designed to mix freely. I've posted a 200 dpi scan, well worth studying.

I really like these ideas, but in a world dominated by Arial and Comic Sans, it's pretty much an uphill battle to get people excited about subtle quality improvements, especially when they require extra time and energy.

There is a lot of precedent for alternate letterforms to aid justification.

Beginning with Gutenberg's 42-line Bible.

Right, but using an alternate character is quite different than electronically adjusting a character's width. The former will look much better and more harmonized in text than the latter.

Yeah, and it would be even better to write a plugin that could justify/balance lines by using the alternate width characters.

using an alternate character is quite different than electronically adjusting a character's width.

It depends how the the adjustment is done. Again, you are probably thinking in terms of universal scaling, but there are more intelligent approaches to glyph width adjustment, which take into account stem structure and weight. The two outer ds in the illistration below have been mechanically adjusted. The one of the left is 14 units narrower than the default glyph in the middle, and the one of the right is 14 units wider (on a 1000 unit em). [This is actually quite extreme for the kind of adjustments we're talking about for justification.] The stem weight remains untouched.

This illustration was made in FontLab using the interpolate function, but the same principle could be applied using, for example, hints to determine stem structures, during the setting of type.

I really appreciate all the thoughts and opinions.

Hrant, that's a most excellent point, and something I've been trying to keep in mind from the start.

I'm currently playing with algorithms for doing visual alignment of margins, automated kerning (especially in the context of maintaining even word densities), adjustment of visual spacing between lines, and avoidance of vertically-aligned glyphs and spaces ("rivers" and related issues). As I write this I'm also wondering if varied glyph sizes could be used to improve the appearance of words with many open letters; probably not, but it's probably worth trying sometime.

I'm sure it's all been tried before, though it sometimes seems people initially try really naive approaches, then, when they don't work well, claim it will always require human intervention and give up on the idea entirely.

To pick on just one particular issue... not that I don't trust font designers with my life or anything, but evidence suggests many aren't consistent about creating good kerning tables. Of the fonts I've been working with, several are missing some rather essential kerns. Rather than complaining about "those blasted inept font people", I believe it's patently unreasonable to expect them to create every possible kern. As time goes on, the demands upon the font engineers to perform similar tasks are increasing. A good visually-based kerning algorithm, coupled with suitable overrides, would be a preferable approach.

I'm happy with the results I'm seeing, but I'm confident much of it is quite subjective--and most of the changes are, indeed, rather subtle. The improvement I've noticed the most is with using visually spaced lines rather than a purely mechanical approach, with glyph scaling and adjustments to inter-letter spacing a close second. I've also been experimenting with optionally using ligatures to improve spacing and word density; this may be helpful, but given the usual set of available ligatures the opportunities for this are rare in 'merican text.

It also seems clear from what I've read and seen that different people have different thresholds for what is most irritating. I almost always find excess space visually disturbing (especially in the midst of an otherwise closely-set text), but other people apparently don't seem to mind it--I say this because it crops up so frequently. Some people obviously loathe scaled glyphs, but in suitable amounts it doesn't bother me in the slightest. Not a few typography authors insist upon fixed mechanical leading, but I seem to prefer using a bit wider leading than the usual recommendations, along with narrowing visually wide spaces (lines with few descenders adjacent to a line with few ascenders). I've seen occasional recommendations for hanging punctuation, but I think without paying attention to visual alignment this usually produces suboptimal results.

I very much agree with John Hudson--simple mechanical scaling may not produce the best results, but there are algorithms which would work better. (But I'm also convinced that, at least at typical text sizes, simple linear scaling is an acceptable compromise for less than 1.5% changes.)

I've placed a sample page of output here (15KB PDF file). I've deliberately toned down the kerning a bit, and of course I chose the font sizes, but otherwise there's been no manual tweaking. I see many things which can be improved, but I think it's a decent beginning. (I just hope I got the PDF protections right; wouldn't want any of those nasty font pirates to grab a few characters. ) The text was a random selection from a Project Gutenberg book.

Raph, that was exactly the sort of stuff I'm looking for. Thanks!

One of Peter Bilak's entries for the Minneapolis-St.Paul typeface competition was a typeface that used OpenType features to set justified text. It had alternate characters that grew wider or narrower, and added or dropped elements (like serifs on the "I") to aid this

One will eventually be able to do stuff like this with the OpenType 'jstf' table, which allows you to set priorities for various actions to assist justification. At present, no systems or applications support this aspect of OpenType, but I believe they will in future. The driving need for this technology will be Arabic justification, i.e. kashida insertion intelligence or other approaches to stretching letterforms.

In the meantime, there is the Justification Alternates layout feature, which provides discretionary, i.e. manual, access to alternate forms intended to assist justification. I'm afraid the only applications I know of that provide a UI function for this feature are the Middle East versions of Adobe apps.

All my Hebrew fonts contain the traditional wide forms used to justify text. Unfortunately, these have fallen out of use of so long in Hebrew typography that I'm having trouble ascertaining what the rules were for their use, presuming of course that there were rules. I'm hoping Nigel at St Bride's will be able to help me with this when I am there in February.

> there are more intelligent approaches to glyph width adjustment

Sadly though none are available on the ground.

> there is the Justification Alternates layout feature

Well, that's not bad. If you make some variants of very frequent letters (like "e" and "t") you could come close to full justification. In fact the "gently justified" approach* where the ends of lines form a nice rag instead of a rigid military line to me seems like the ideal general objective, and variants of some key glyphs would be plenty for that.

* Which the Ancient Greeks used but virtually
nobody today does, except for things like
email! Well, at least I do... :-)

--

> I'm sure it's all been tried before

Don't be so sure.
It's pretty rare that coders have been guided by the pragmatic, fuzzy world of good typography.

> it's patently unreasonable to expect [type designers] to create every possible kern.

It's also not a good idea, both technically as well as in terms of cost-effectiveness.
Which doesn't mean however that most designers do an adequate job; typically it's a process that most resembles the path taken by a headless chicken; methodology not being a forte among visual people, good spacing/kerning is the first thing that gets cut.

42nd, this stuff you're working on seems promising, and trust me it's needed!

> One of Peter Bilak's entries for the Minneapolis-St.Paul typeface
> competition ... had alternate characters that grew wider or narrower

But of course that wasn't cute enough.

hhp

what can't opentype do?

Be implemented fast enough.

> Do people just think that it "looks better"?

Modernism wants things to line up. It doesn't like variety. The problem
started with the Romans, and really took hold during the Enlightenment.
There's even a Monotheism connection in my mind. It's really the biggest
problem with Europe, otherwise it's a great place to live/be_from.

OpenType:
There are a lot more things it could do (like linguistically-
aware glyph selection) if applications would only support them.

hhp

Hrant, I really appreciate the feedback, and I shall continue working with redoubled fervor.

I think rigid justification has typically been used because people got into the "one or the other" style of thinking--i.e. either you fully justify or you do nothing at all, and the do nothing approach with a very uneven rag often looks worse than rigidly-justified text with holes in it. When justification was "new" it probably added a very impressive look to a document, so nowdays it is de rigeur for most publications.

It's also harder to do, so it obviously must look better. I say that not as a jab at anyone in particular... but too often we believe that because something is difficult it also must be worthwhile.

I'd bet Eric Gill's book of typography essays has successfully discouraged at least a few typographers from using ragged-right justification. The rag he used was "natural" but unattractive, and there were any number of places he could've done better but opted not to.

And... people find symmetry attractive. (That seems to be one of the few universal constants when it comes to beauty, most especially with human faces.) A document with rigid justification on both sides can be seen as being more symmetrical than one without.

I've been meaning to look into all these fancy OpenType features... now maybe I have an excuse to.

> too often we believe that because something is difficult it also must be worthwhile.

Yup.
While Nasa was spending millions to make a special synthetic material to tile the surface of its orbit re-entry vehicles (which apparently still manage to blow up now and then), Soviet space enigineers used a material which would simply burn off: wood. But of course bosses make more money when resources are wasted.

> people find symmetry attractive.

This is admittedly true*.
Although the human face is actually subtley asymmetrical, although subvisibly so.

* And as I pointed out elsewhere this works against [conventional] stroke ductus, like in the "A".

hhp

John,

If you talk to a 'sofer', a Torah scribe, he will be able to tell you the rules in detail, I bet. I have the idea that the wide characters are typically used at the end of a line. A scribe can vary the width of the letter to fill out the line, of course. I suppose this could be programmed also, but also modern eyes may find it a bit distracting.

Also about rubber type in general, Fellici recommends only a +/- 2% stretching, and says it does help readability. I tried it with Minion, and didn't find it bad. But I haven't made a systematic comparison.

Yes, I will talk to a sofer about this, William, although I suspect the rules might be more strict for Torah scrolls than elsewhere. I've seen a variety of uses of wide letters in early printed books -- and not all at the end of lines -- , but have not yet taken the time to confirm whether there are general principles governing their use.

>I suspect the rules might be more strict

These may be dictated in certain cases. I know some large letters, bent letters etc are demanded for various reasons here and there. But overall, I would expect that it reflects scribal tradition. What is neat about the Torah scribes is that they are living representatives of an ancient skill, with a continuous and unbroken line from antiquity. And they are quite different from calligraphers, to my mind, as their concerns are more like typography, as they have to do a lot of text that is readable and attractive.

I had occassion to speak to one sofer and he told me that the hardest part was that each stroke of each letter had to be done with 'irat shamaim', the awe of Heaven. Talk about your demanding clients!

A sofer in the UK has a fascinating website here--well worth checking out.

Yes, that's an excellent site. It has been very useful in my research over the past couple of years, especially regarding nun hafukha.

When I first read about inverted nuns, I had momentary visions of The Flying Nun...