A new approach to spacing?

Ramiro Espinoza

@Terminal Design I just opened a random unfinished font to make this screenshot. The font only has a very quick automatic kerning, so yes, it's not good. 

John Hudson

What I want is a font to have optimal spacing/kerning.

It's worth taking a moment to consider what this means, given that, at least in my experience, a font is a particular instance of a typeface and seldom represents a final statement. Typefaces get extended or modified in various ways, and for many of my clients this has to be done in such a way that the resulting new font does not break compatibility with the existing version(s). Most typically, clients don't want any changes to existing spacing/kerning outcomes, which would result in reflow of existing static documents.

So, for me, 'optimal' involves factors beyond most efficient data expression. Optimal spacing/kerning has to be easily extensible, and to be so without changing any outcomes for existing spacing/kerning. So re-running an autospacing tool on a new build of a font, would have to either a) match exactly the spacing/kerning data for existing glyphs in a previous version, i.e. not letting that data be influenced by the inclusion of new shapes, or b) functionally match that data in terms of spacing outcomes.

Ben Blom

For an auto-spacing tool, it is indeed very important, that it would support this: Update the spacing/kerning of an existing font which has been updated (by adding glyphs, and/or by changing the shape of glyphs that existed in the old version), in such a way, that the definitive updated font does not break compatibility with the old version of that font (preventing reflow in existing static documents).

What would be a practical way to implement this? Open both the old version and the updated version of the font in the auto-spacing tool. Also open a file that contains the exact settings (both the relevant parameters/options, and any manually set distances between glyph pairs) which were used by the tool when the old version of the font has been spaced/kerned. In this way, the tool knows both the spacing/kerning of the old version of the font, and the “optimization logic” which has been used when creating the spacing/kerning for that font. (Implication: the tool should create such a settings file, whenever it spaces/kerns a font.)

When updating the spacing/kerning of the updated font, the tool should keep the existing spacing/kerning of [distances between] all the glyphs of the old version of the font, and only add spacing/kerning for the added glyphs. One exception to this: If the shape of glyphs that existed in the old version of the font, has got wider or narrower in the updated version, then the tool should update the sidebearings of such glyphs—for instance, by subtracting half of the change in the width of such a shape, from both its left and right sidebearings. (Perhaps I wouldn’t use the word “optimal” for such an update, but would describe it as “not changing existing functionality”.)

Nick Shinn

I don’t believe there is such a thing as “optimal” spacing or kerning.
It’s a matter of style, dependent on such things as the typographer’s taste and the document context.

When I’ve revisited my type designs after 10–15 years to update them for OpenType, I’ve found that another person apparently spaced and kerned them.

When I was using Quark in the early ’90s, I did a lot of marketing for a bank which used Palatino as its corporate face, and I re-configured it (without changing the font), by slightly condensing and tightening it up (in the style formats), and changing the kerning tables (a facility of XPress).

Certainly, auto-spacing /kerning tools can help type makers, but there is also a possibility for them to be layout algorithms for type users. At the moment, Adobe’s Optical Spacing has only one flavor, but I see no reason why “Tight But Not Touching”, for instance, shouldn’t be an option. Here c.1970 Carl Brett used the default spacing in the smaller Helvetica type, but hand-kerned the title in TBNT style. (Yeah yeah, I know, many of you will think it looks like crap, but my point is entirely philosophical!) So either the foundry can apply a different spacing/kerning model to display cuts (which I have done occasionally), or an algorithm could be used in layout by the typographer, perhaps as a plug-in.

Chris Lozos

Some of us geezers loved the tight but not touching era but that "Poetry of Relevance" example is not a good one.  That "ry" rumps the whole thing and the "va" just ain't makin' it.  ;-)

Ben Blom

I don’t believe there is such a thing as “optimal” spacing or kerning.
It’s a matter of style, dependent on such things as the typographer’s taste and the document context.

“Optimal” doesn’t mean “best”, “esthetically right”, “good”, “what someone prefers”, “the only way one should do it”, or something like that. It is a “technical” term, describing the result of an optimization process, which optimizes something (like minimizing the number of kern pairs), given the value of parameters, the position of sliders, the value of options, any constraints (like manually set distances of key glyph pairs), etc.—and, of course, the shapes of the glyphs of the font involved. Perhaps you have missed the “given...” part of this description, because such parameters and constraints are meant to set any “variation of spacing”, or “style”, or “preference”, or “handling of particular shape relationships”, or whatever, that the designer wants. So this is not about one flavor, but about a multitude of flavors. The number of optimums may be practically endless. One way of setting the parameters/options of an auto-spacing tool, should result in “Tight But Not Touching”—and there may even be variations of a “Tight But Not Touching” setting. Of course, a challenge for the maker of such a tool, may be to get all the relevant variations of spacing, style, etc., in the parameters/options/sliders of the tool.

It is in interesting idea to give a “multitude of flavors spacing power” to type users as a layout algorithm—if this would be possible. It may be a good thing for them, because it gets them more control. However, type designers might protest, because they may feel they would lose control of setting the “spacing style” of their fonts.

Chris Lozos

"there may even be variations of a “Tight But Not Touching” setting"

There certainly are, as Nick and I just demonstrated.  My point is that it is very hard to program an aesthetic and get agreement on how well it works.

Kent Lew

because they may feel they would lose control of setting the “spacing style” of their fonts.

“Control” is an illusion. Type designers are not in control. We do our best, according to our lights, and then we turn it over to the world. Typographers (and now others) will do what they will do. Including turning on Adobe’s “optical” metrics.

Scott-Martin Kosofsky

I believe there is a kind of “classical balance” that the best-made text types have, a balance of internal and external space that permits a slightly tighter or looser spacing within a reasonable range of sizes. But to achieve that, you have to understand what you cannot do: you cannot make a font that works for text and also accommodates TBNT spacing. The “Poetry of Relevance” example makes it clear why that is so. The TBNT style was a byproduct of photo headline setters of the 1960s, which were manually operated to set letter by letter. The operator never saw the entire word until the last letter was set. To achieve a smooth result, the letters were cut apart and spaced in paste-up, from which a final version produced by photostat. (Anyone remember “stats”?)

Perhaps, some day, one will be able to make fonts that include more than one kerning set, perhaps as a novel alternate under “Sylistic Sets” or some other feature that has yet to be defined. But why bother? It would be much easier to keep the text type for text and make a display version that’s spaced differently. The fact that the spacing for text and the spacing for display are inherently different things makes the idea of an automated spacing tool unlikely to work well, unless the tool includes a variety of settings.

Nick Shinn

Including turning on Adobe’s “optical” metrics.

More than once, I’ve had clients who commissioned a typeface and given feedback after they’d tested it in their workflow—in which “Optical” was the default. Such faith in algorithms, just push a button and everything will be perfect!

**

I have had occasion to use “Optical”, it’s quite fun to apply it to monospaced fonts.

**

I spaced Richler quite openly, and a bit strangely for an oldstyle, and suggested that if typographers decided to use what is mainly a body text face, for display, they might prefer “Optical”. 

Chris Lozos

As type designers, we have to give just do to our users.  They are the ones who have to make our type work in their situations.  This does not mean they will always make the same choices we would.  This does not mean that some of them won't make an absolute mess of it.  We have to allow them to do their job to the best of their abilities [whatever they may be].  We are in the business of providing tools, not total out-of-the-box solutions.  I am sure a Roger Black would do a fabulous job with any type but some kid who only took a class in InD last week and has no design experience would botch even the best type.  I think we have to design for the competent type user and allow the not so competent ones to learn as they go.

Even the "Optical" setting has its flaws.  I would rather fix it by hand than trust whatever "Optical" does.

[Deleted User]

The user and all related content has been deleted.

Chris Lozos

Franklin Gothic needs alot of help ;-)