Caps and lowercase spacing

[k.l.]

Bill – Kindersley believed that once you had correctly positioned a glyph between its side bearings, and had the total advance width correct in relationship to other glyphs, then tracking should make no difference.

The middle part, and had the total advance width correct in relationship to other glyphs, is the crucial one and, my impression, is an aspect that Kindersley dealt with rather silently in his approach, possibly didn't even care about. His article "Optical Letter Spacing" (Penrose Annual, volume 62, 1969, pp 167–176) shows, on p 175, a comparison of tight (left) and wide (right) spacing as created with his approach, by example of the caps alphabets of Folio, Times, Perpetua and Optima. The tight spacing column is telling: The L is too close to, and sometimes even touches, the M which follows. He doesn't mention this explicitly yet this demo shows that overall spacing width is an implicit factor. If too tight, the method does not yield proper results.

Could you explain more what you mean by "contextual positioning?"

It is how I translate my initial observation, that caps need one kind of spacing when in all-caps context and another kind of spacing when in lowercase context, into the current spacing-plus-kerning paradigm and OpenType add-on: There is one set of sidebearings for lowercase context; this is the default sidebearings. Then there is another set of sidebearings for all-caps context; these 'alternative' sidebearings are implemented by adjusting glyphs' positioning and/or advance widths contextually, via the GPOS table. (This is what I did in some of my fonts. Another factor, though, is each typeface's design. Not every typeface needs caps to be spaced differently for different contexts – a number of recent typefaces have rather narrow caps which produce the same 'beat' as lowercase letters anyway.)
Indeed 'normal' kerning can be interpreted as a special case of contextually adjusting positioning and/or advance width.
Kerning triplets sound interesting for dealing with a few exception cases but is not what I was thinking of.

The approach I followed was as you say trying to cope with our existing system. ... But I was also concerned with the practical problem that you don't want your font totally wrecked with both kerning and contextual alternatives not used.

And this makes pretty much sense. When I wrote that my comment is not to suggest that your implementation approach is wrong, this was not mere rhetorics. Over time I got weary of trying to push the boundaries of the current paradigm within the current paradigm. It only gets one so far. I realized that one better acknowledges a given technology's limitations – or come up with new technology. And, like you say, it looks like no one else gave a shit. :(

hhp – BTW Karsten: All good, just no "rhythm" involved.

I guess you aren't surprised that I am not surprised reading this. ;-)

[hrant]

You might however be surprised to hear that I wouldn't
have said anything if William hadn't repeated the term. :-)

hhp

[k.l.]

I am!

A little update. I wrote that the overall spacing width ('beat') is an aspect that Kindersley dealt with rather silently in his approach. In the Cardozo Kindersley workshop's "Optical letter spacing for new printing systems" publication (2001 edition) he does mention it when discussing the OILIO test string: For practical reasons kerning is undesirable in the initial stages of dtermining relative optical spaces for characters. Therefore close setting which would produce this is to be avoided. ... The ideal working set therefore would be when the 'L' is provided with the minimum space without any part of it projecting beyond its space. (p 31) Though the description of his approach begins with the spacing of O and I, actually it is the L which determines the 'beat' for spacing O and I. (The dramaturgical order doesn't necessarily mirror the order of what the story is about, true for fiction as well as non-fiction.)

[William Berkson]

Karsten, I was aware of the idea of initially determining looseness or tightness with the L. But somewhere he also says that overlapping is preferable to changing side bearings. so the idea of further kerning seems to be something added after the death of Kindersley. But I may have the story wrong, and I don't have time to look back over it. In the end, I think Kindersley's idea of an optical center is good, but his ways of determining it—other than by eye—never really worked. And he didn't acknowledge that at tighter settings kerning involves compromise between different ideals, such as even color vs not having letters touching.

[k.l.]

Yes.

[Nick Shinn]

I didn’t consciously develop a method for spacing Parity (unicase).
It struck me that it has more of the qualities of minuscule setting when set tight, and the qualities of all-cap setting when letterspaced. That may have something to do with the fact that it’s a lining style.

Another thing with Parity; as it is configured as “Unicase with Small Unicase”, I was forced to be more specific and describe letters as majuscule, minuscule, or common, rather than in terms of “Upper and Lower Case”.

Anyway, I think I have acquired some extra knowledge about spacing through working on this style, although I’m not sure how it will play out in my next bicameral design.

I try not to systematize things like letter spacing; I don't think spacing is something to be added on after glyph shaping—there’s a feedback loop in which glyph shape and spacing evolve in sync during the design process.

Some ideas I’ve tried: wider sidebearings for “I”; narrower serifs for letters such as “E”; different serif widths at top and bottom.

Slab serif types are the worst to space without kerning.

[dezcom]

In order to avoid very odd first glyph in line spacing for caps, I set side bearings to fit better with lowercase [tighter] and kern positive between some caps classes. This avoids the need to use capital spacing feature.