f T isn't a ligature - it's simply a small word space. Perhaps the space is kerned towards the f and the T towards the space, maybe it's just an unfortunate character combination.

A Tale of Two Cities (classic problematic kerning setting).

Until three-character kerning is implemented in fonts, and supported in layout apps, you will have to manually apply “Optical” kerning.
Perhaps there is some way to code specific exceptions?

Actually, now that I think about it, it would be possible to solve this instance in an OpenType font:
sub f space T by fTspecial;

…where fTspecial is one glyph.

Done:

I might not be getting something here, but this doesn't need a fancy three-letter super kern if the 2 pairs are present... This kind of thing is precisely why I'd argue it's a Good Idea™ to kern things like f/space and space/T. And maybe Q/space and space/J, depending on the design. Overhanging glyph parts that optically diminish the space. Even if an f/space combination isn't met with yet another overhanging thing on the other side of the space (a not-too-unlikely worst-case scenario), it can still optically make the space appear smaller than it should be.

Thanks for the replies… I have found a post-script version of Adobe Univers and it doesn't have any of these problems so I'm switching. It seems odd that the more up-to-date open type version has these problems?

What, they have negative kerning in pairs like f/space and space/T? Why?

Univers has space/T kerning -180, about 2/3 of a space (278), hence the problem. -20 or -30 is quite good.
BTW in John Hudson's Brill: f/space +25 and space/T -20, and looks like the best solution.

William, I understand the practical issues, but
shortening the beak of the "f" before a space is
pretty much opposite of what should be done!

hhp

I think a minimum (apparent) wordspace is important,
but if wordspace variance were that bad justified setting
would have far fewer proponents. I think boumas are
more important.

hhp

Maybe this depends [too] much on spacing (and design) to formulate a general rule.
If the "f" already has a negative sidebearing, I can't really imagine it would need a negative kern on top of that.

FWIW, I personally don't see things like "T"s as having "a hole under" them. The x-ht region isn't the only thing one needs to consider when kerning, is it?

Here's the illustration. Top is what CALT gives you. Middle is with the default long-nosed f, and the negative kerning of the space and the left side of the T. Bottom the negative kerning of the left side of the T to the space is removed. The other kerning (Ta, Tw) is retained.

To me the top looks better. In the middle the closeness of the f T is distracting. In the bottom, without the negative kerning of the space and T, the relevant word spaces seem a bit wide.

@Kent: Maybe you were just being facetious, Nick.

No, before I posted I considered that this “ligature” would default to multiple glyphs once beyond a certain tracking value.

I also like the suggestions made by Chris and Bill, to substitute alternate “f” or “space” glyphs.

**

Paul, how does one implement contextual kerning triplets in a font?

Nick: No, before I posted I considered that this “ligature” would default to multiple glyphs once beyond a certain tracking value.

Hmm. Interesting assumption. I’m not sure that the adjustment of word spaces by the justification engine would trigger the same mechanism that tracking does, such that an f_space_T {liga} substitution would be disassembled in a looser line. But maybe you’re right. One would have to test this.

@Kent: I’m not sure that the adjustment of word spaces by the justification engine would trigger the same mechanism that tracking does … One would have to test this.

Good point, I didn’t think of that.

How much is "maa" compared to "a hair"?

hhp

Urban Dictionary is inconclusive.
I therefore propose that "maa" is between "a hair" and "slightly".

hhp

@Bill: … odd or antiquarian.

That’s me alright, I’m afraid I much prefer the originals to Linotype’s makeovers—but good for Frutiger and Zapf to be making a buck off them.

Nick, there is no mechanism to add kerning triplets (or quadruplets, or... I’ll just call it contextual kerning) when exporting from FontLab, but if you’re compiling your fonts with AFDKO, then the syntax would look something like this:

pos glyph1' 22 glyph2 glyph3;

I might not be getting something here, but this doesn't need a fancy three-letter super kern if the 2 pairs are present... This kind of thing is precisely why I'd argue it's a Good Idea™ to kern things like f/space and space/T.

It's a bad idea. If you're going to use kerns with spaces, as Nick says, you need three letters. You can never anticipate what will fall together. One of my test words in my test-text checks "f_space_J" as with "of Jewish" You use a pull-back kerning with the f, and the space can disappear.

Or look at Adobe Jenson, with the f_space _quotedblleft. Again, the space disappears -- really, it usually looks like it falls after the qutedblleft...

Better to use an alternate f when followed by a space. Take part of the space out of terminal of the f, and add a little to the sidebearing.

A lot also depends on how tightly you set type. If it's set as tightly as Kinross's books, you'll find all sorts of oddities with the spaceband. In Hochuli's book, Detail in Typography, he argues for a certain generosity with the sidebearing space around parenthesis, quote marks, brackets, etc. (p. 37) . Good point. But with tight setting, you'll find that space sometimes greater than the wordspace. There isn't an example in the text of DT, but you can find plenty in Dowding's Finer points.

Remember, back in the days of foundry type, the comp just didn't add space -- or as much space, when optical considerations were made, when driving out the line to justify. To get a computer do have the same effect is a lot of work. If it's ever published, I spent a good bit of time on this in my chapter of Rich Hendel's new book. Everyone is encouraged to buy a copy (Shameles plug) (Hey, Kent has a chapter in it, too) (if it's ever published). (Except Hrant, who insists he'll get it from the library.) (And no, I get no royalties -- it's Rich's book).

Ah. You remember my hobby.

"InDesign supports contextual kerning triplets" Paul

A-space-T and f-space-T are triplets, first -100, second +300 (/2000)
Lower row - kerning is set to zero

Hrant: 'what happens to regular (double) kern values that might also be specified?'

"Contextual positioning rules must be in a different lookup than pair positioning rules, since the rules are of different lookup types. Because each lookup is applied independently of the other lookup(s) over the entire text stream, the positioning change specified in a pair kerning rule will be added to the positioning change specified in a contextual kerning rule, whenever the two rules match the same glyph pair in the text stream."
[ http://www.adobe.com/devnet/opentype/afdko/topic_feature_file_syntax.htm... ]

Hrant: 'In a triplet, is the value applied between the last two?'
examples from cited site:
position quoteleft T' <0 0 20 0 > quoteright;
This form shows a full value record which allows you to alter both the (x,y) coordinates of the origin and the (x,y) coordinates of the advance width. So in this example the advance width of "T" will increase by 20 (the origin remain unaltered).

position quoteleft T' 20 quoteright;
The simple form of the value record which specifies a value for only a change to the (x) value of the advance width."
So it is your choice: you can shift the glyph and/or change its advance width.

And the Contextual Kerning is not limited to two or three glyphs:
position s f' 10 t period;

HaHa! Got it! You can do it with probably many glyphs/classes, as far as you keep them in three groups and one glyph/class (middle) is affected:
pos o f space' T w o 300;
or even more than one glyph/class:
position s f'10 t' -5 period;
where in sequence s-f-t-. the x-advance of f is increased by 10, and the x-advance of t is decreased by 5.

The upper text matches the rule, the lower not, so only first space is wider.
[support in AFDKO added Jan 2009]

POC:
lookup kern1 {
pos \s \f' 200 \t' 100 \period ;
} kern1;
feature kern {
script DFLT;
language dflt ;
lookup kern1;
} kern;

Update: "start" and "end" groups aren't mandatory. You can do something like this:
pos \a' <100 0 200 0> \b' <100 0 200 0> \c' <100 0 200 0> \d' <100 0 200 0>;
at every occurrence of the string "abcd" the letters "a b c d" are extra spaced (in this example no matter what is before and after).
(tested in FireFox 14.0.1; pair kerning is still additive)