Each pair can only be defined once for an assumed average context, which really only does the basic minimum for you, and is fine for body text. But now let's say you're designing a logotype or book cover. The overall context of any specific setting can change how letters ought to be kerned so that the whole word looks consistently spaced. If your design calls for really tight letterspacing, every letter may best be kerned from scratch.

Imagine you've done a bold logotype for a company called INNERHIM. Nice straight sides you can shove together like books on a shelf, and the extra space made by the R's tail is acceptable because it falls at a semantic division (You would probably go for an overlap if the same RH were in, say, MINIRHINO). Now the company says it wants the same look but its name has been changed to OVERLACKY. Can you at least leave ER the way it was? No, because there is so much necessary space between all the other letters, ER will look too cramped if it is not loosened to match. This is an extreme example for the sake of explanation, but smaller decisions of a similar kind have to be made all the time.

Thanks.

Kerning from scratch...
I always stay uncertain when doing this, never know when to stop or where to rely on. I get stuck comparing spaces over and over making futile changes even when it looks pretty decent.
At the moment I use the character spacing rule of thumb as described in 'Letters of credit' but I guess I'm looking for THE method / system.

…anal people…

Sure, my fonts are well kerned—I only wish the rest of my life was so neat and tidy!

**

Yes, manual kerning is a matter of taste.
However, the style of (headline) kerning which requires least deliberation is “Tight But Not Touching”.

Spacing/Kerning is a very zen thing. Embrace the uncertainty. And when you get zen about the black bodies of the letterforms as well, that's when you've fully moved up from lettering to actual type design.

hhp

...the style of (headline) kerning which requires least deliberation is “Tight But Not Touching”...

Wouldn't that be a good / easy starting point then and take it from there using tracking to loosen things up if you're not looking for tight?

...Embrace the uncertainty...

I guess that is the hardest thing for me.
Is there really no science to it?

I guess that is the hardest thing for me.

Same here. We've been brought up Modernists, and it's hard to see beyond the illusion of Control.

Is there really no science to it?

Some. But there is Art even in Science.

hhp

...the style of (headline) kerning which requires least deliberation is “Tight But Not Touching”...

@Cor: Wouldn't that be a good / easy starting point then and take it from there using tracking to loosen things up if you're not looking for tight?

Not necessarily, because it’s a totally different spacing system, for headlines—as opposed to Tracy’s method, which is for body text and book faces.

However, I have used the method you suggested, for instance in Figgins Sans:

Quark XPress enables editing of tracking according to size, and also custom kerning of fonts. However, it does not enable varying kerning according to size.

Even in text fonts kerning round-round pairs positively can be a good idea.

hhp

...However, I have used the method you suggested...

Thanks.
Just to make sure, you used Tracy's method to kern the upper example then applied -30 tracking to get the lower example?

No, I used my own method to get the lower sample, and then opened it up for the published version of the font.

Ok gotcha!

Thanks Nick.

Lately I've been reading a lot of articles on kerning and I've been reading a lot about type designers and their meticulous ways.

By the way, if you go back to the Dark Ages -- before, say, 1992, programmatic kerning was done by compositors. For years, Linotype gave as one of their reasons for sticking with the 18-unit em for kerning & I believe tracking values was their customers wouldn't want to change their kerning programs. Their customers, of course, were all typesetting shops. As if we couldn't multiply by 3... (Yup, the new em was a 54-unit em. Don't remember the year - late 1970s? -- Nick Shinn would probably know.)

To this day, except for Matthew Carter's fonts, if a font's primary use in our shop is to be for setting text, the first thing I do is throw out the foundry kerning. Matthew's kerning is almost always to my taste, just not quite extensive enough, esp. with punctuation.

There is a point to this nostalgia. There is pretty much no way a type designer can get all the needed kerning, and absolutely no way they can accommodate all legitimate differences of taste on "proper" kerning. Don't blame them if things aren't to your taste.

Typesetters are not people who only know what the "place" command is, and can fill out paragraph styles. They have other responsibilities. In the end, proper kerning is the user's job.

Ah, you're a guy who prefers libre fonts, right? Maybe you've been burned too often by too many lousy fonts...

I hope one of my babies never ends up in your bathwater - people will look at the setting and think I suck at kerning.

hhp

Ah, you're a guy who prefers libre fonts, right?

No, I'm one of the guys who prefers good type, where ever you find it. If I just wanted to be nasty (as seems your favorite activity), I'd say I hope none of your fonts winds up in our font library, too. But since to the best of my knowledge I've never seen any of your fonts, I have no reason to say that.

I'm waiting for this day:

Newsflash
Abobe just announces that (doesn't matter, say Arno) will become open source. Upon the announcement, Hrant H Papazian immediately reversed his long-standing opinion that it was an excellent font...

Wrong again - I think Source Sans is great*. But you have to admit, good spacing is generally the biggest problem when using a free font. Like check out the delusions of kerning grandeur in this gem:
http://vissol.co.uk/mavenpro/

* http://typophile.com/node/95313#comment-518000
http://typophile.com/node/95280#comment-518303

And "nasty" is throwing out the kerning unless Carter made it. Actually I can think of some other adjectives but I'll be nice and keep them to myself.

--

BTW, why did you remove the "2,000+" from the bit about your font library?...

hhp

Try this one from Adobe -- At least, as late as 2009 version of Warnock Pro, which probably came free with some software you've got...

set a single open quote (aka quoteleft), a capital A, a single close quote (aka quoteright), and a comma.

Let us know how you like it.

With Adobe Jenson, set an f, a space, a quotedblleft, and something else -- say a lower-case l.

Let us know how you like it -- hint: it's the visually transposed wordspace.

Why I throw out kerning. To many surprises that can ruin your day if you discover them when first proof is due in three more days...

Edit:

I removed the "2,000 fonts" because it seemed extraneous (I don't write well, too prolix). And it's a bit illusory. We do have over 2,000 fonts, but not that many families. To use the number seemed give the wrong impression -- as well as being irrelevant.

In Minion, "Yp" is kerned but "Yq" (which needs it more) isn't. They're about equally [in]frequent: Ypres and Yqem (AFAIK the most frequent words exhibiting those pairs) are both relatively obscure Western European localities. The former is losing relevance as WWI history becomes more boring; the latter is gaining relevance as wine becomes more interesting. My Mana series of bitmap fonts has both those kerned. I average ~8 kerns per glyph, all hand-made - no kerning classes (most recently for a reputable medium-sized foundry that paid me handsomely). Caveat: I have yet to deal with very large charsets.

What makes you think I consider Adobe fonts the ne plus ultra of spacing (or anything else)? But they're still so much better than 99% of the open/free/libre stuff. Because somebody paid somebody good money. AND Adobe allows modification, so precious to you and me.

BUT, all that said: You want perfection? You gotta leave this solar system. We don't do perfection 'round these parts.

Why I throw out kerning. To many surprises

You want to reduce surprises? Throw out technology.

hhp

But they're still so much better than 99% of the open/free/libre stuff. Because somebody paid somebody good money.

We're ranging a bit far afield here, but on that model, politicians who take graft are then better than their more honest counterparts.

* * *

And don't forget Yg -- as in Ygraine, King Arthur's mom...

Except that in a democracy an honest politician is an imminent failure anyway.

hhp

Actually I now realize that -at least online- Ypsilanti is more common than Ypres (and -more significantly- Yqem) although its relative historical insignificance means it wouldn't come up much in books and stuff. :-)

hhp