Oh goody! :D
Thanks for the heads up, Igino! Just downloaded the opentype versions. :)

Let me take this opportunity to tell you something: I have to say I LOVE these. I've loved them since I first layed eyes on them in 2006, according to the date on the folder where I religiously kept them. In fact, I have no religion, so this might be exactly what I mean, literally; type is probably the closest thing I have to a religion. :P
Anyway, I cannot thank you enough for sharing this amazing work freely. The Fell types really were one of the reasons why I started loving type and wanting to create type in the first place. You rock!

Ruben

DITTO!!!

YOU ROCK INDEED!!!!!!!!!!!

Igino, what is your rationale for kerning a facsimile-style revival of an old metal type?

Is IKern a program you sell?

How does it compare with KernMaster from DTL?

Nick: He says on his page

"First of all: Love. Love for beauty of this typefaces I'm still not tired to look at. I begun to learn how to build a typeface just after having seen them because I just wanted to use them. And after having digitized them I begun to think about a way to space them (it was so difficult): the seminal thoughts about iKern, the autospacing and autokerning tool I've developed since 2002. I can certainly say that without the Fell Types iKern wouldn't exist: Their typographic “errors” (variable serifs from letter to letter, even inside the same letter, inconsistent heights and slant angles, different weights and many other amenities and lovely strangeties) have pushed me to a deep generalization of the mathemathical model avoiding shortcuts and suggesting paths I wouldn't thought otherwise.

For this reason I decided to license the fonts for free. I've received enough. And more: designing typefaces is not my job and, in the end, I didn't design the Fell Types. Peter de Walpergen, Claude Garamond, Robert Granjon did."

That said: I would like to hear more about why if he has more to say. Also, you could voice your critique of the idea.

How does it compare to kerning by a human?

I've come to the conclusion that spacing and kerning are absolutely not a matter of Art. They're just a Tecnique.

You could say that about anything! Like playing the guitar, acting, cooking, painting…

—K

I am in Kris' camp on this. Doing things by machine is fine - only as long as a person comes through afterwards. And then only if they actually look at it quite carefully.

Not everybody who plays the guitar can play the guitar.

For the beginner it may be unbelievable how someone has achieved good spacing. It takes considerable time and a lot of attempts and mistakes to get an idea of how to do it, and maybe master it. Still this is more or less "intuitive", unconscious: being good in doing something is not the same as understanding what one does. The next step is to analyze what the master does and formulate a method. In so far, being creative and creating a machine that does what a creative does is not completely unrelated. And of course once a task is automated, this raises the question, was this task creative at all?

From a practical point of view, I do not see why someone should spend days on a task which can as well and with virtually same results be performed by a machine. Would there be a loss if spacing were done by a machine rather by a human being? I cannot tell you a single typeface of which I could say that its spacing exhibits its designer's creativity. Rather, some typefaces work, some don't. (In terms of spacing.) The question is not: is spacing/kerning art or technique? but rather: is the applied method ok? It is there that different type designers may have different preferences -- provided that they would ever transform what they merely do into a clear conception!

And we should always keep in mind that Gutenberg's one-letter-per-block, one-block-next-to-another paradigm of typography (which implies that the space between two letters is defined by the remaining side bearings) was more or less a necessary evil of this very physical approach to letters/type. Letters an sich do not know side bearings nor kerning, they only know the space inbetween. The real scandal is that digital typography never even dared free itself from merely imitating the one-letter-per-block paradigm. Despite the fact that the digital domain knows no physical constrains. The work of Mr Marini and others could finally free it from outdated paradigms. (Why output sidebearing and kerning tables at all?) Again, what is to be discussed are method and result, but not man vs machine ...

To put an even finer point on it, many discussions have arisen here about the role of "white space", which is exactly what Inigo's program is concerned with. Evading the crucial lessons of spacing sounds initially like a terrible loss in a field where visual discernment has always been crucial. However, this program is not the first attempt to automate the process of spacing and kerning. Though the focus was different, the h-z algorithm Zapf helped develop was intended to produce ideal spacing and kerning of text blocks without direct human manipulation. Though the spacing of a typeface is ideally done by a careful, perceptive human, the presence of a human designer does not guarantee ideal spacing. We've all wished for modules to help us through boring and repetitive tasks in type design, and our software and workflows reflect this. Type design is as much technology as it is art.

I also think Inigo has learned some very interesting things in his experience with the Fell types, things we ought not brush off. Many of us have not had the experience of dealing with highly un-systematic types, though we all appreciate variations and irregularities. The fact is that it is easier than ever for typefaces to be sterile, modular, regular and coldly rhythmic. Cut-and-paste has become a Modus Operandi for too many. I would be very interested to know what the algorithms in iKern are doing and what variables they consider. It might be instructive; might help us remember what to look for.

To William:

> Is IKern a program you sell?

Not anymore.

>How does it compare with KernMaster from DTL?

I don't know. Never had the chance to try it! Anyway iKern does autospacing and autokerning at the same time allowing a straight and productive workflow from the design to the font. About the quality I'm confident the output is, at least, adequate. But everyone can judge it by himself. The autokerning stuff rely sometimes upon informations extracted during the autospacing phase. That's why I tend to consider the “only autokerning” a nonsense and I've abandoned three years ago an “only autokerning” iKern version .

To Kris:

> How does it compare to kerning by a human?

>> I’ve come to the conclusion that spacing and kerning are
>> absolutely not a matter of Art. They’re just a Tecnique.

> You could say that about anything! Like playing the
> guitar, acting, cooking, painting…

How, I'm very glad you give me the real occasion to say many things. First of all I've to say that for years I have surgically examinated supposedly professional fonts. My conclusion is that only few humans actually are able ro really kern (and space). I'd have to come to the conclusion that a tool like iKern should have suited the needs of the rest of the World. But I've been told that the font industry survive even selling fonts without kerning (and the customers don't even notice!). That said for six years I've tried to analyse the essence of this job and I think that what the few humans do is, in very short terms:
fully understand the underlying rules of fitting;
have the capacity, the strength, the time and the will to apply in a coherent and complete manner these rules through all the fonts.
I tend to consider this second part of the action more as food for computer. That's why I always heard from anyone involved that kerning and spacing is not, at least, funny. It's normal. We hate repetitive tasks. We tend to be distracted. We like open air. We like many other things.
I tend to consider the “full understanding of the rules of fitting” a common quality in many people and not the exclusive gift of a religious caste. And this is the real core question. I don't know at which level the full understanding is. Certainly not in a formal, I would say conscious, level. Otherwise a real usable fitting theory would explicity exist today and we wouldn't here to discuss. What I'm saying talking about iKern is that I think to have found a mathematical model that describes those “underlying rules of fitting”. Am I right? Am I wrong? I'm confident to be right when I'm satisfated with the output iKern produces. I'm not confident when I see something I dislike. But I'm there to make it better. Consider I've sometimes re-elaborated the model just to change a kerning value of two points (I admit it. I'm a geek). But first of all I am confident when other people tell me the output is good, sometimes very good, for some enthusiast perfect. And, as the font industry man would say, I'm confident when I find people willing to pay for this service.
To finally answer I would say that there is no difference between what a human does and what iKern does. But to use myself a metaphor: I'm able to walk. I like it. But I'm often use a car. And a GPS too. And more: the model is not strict. Since autospacing has to be done the whole rhythm has to be reconstructed. And the man (me) says how.

Let's now discuss the “Tecnique” side. I'm actually a good guitar player. I had the fortune to have parents that forced me to study music. The piano seemed to me too much demanding so I fell back on the guitar (girls like guitarists. You know). But I don't consider myself a real guitarist. Because I don't have the music in my heart. I can only mimic. I think that spacing and kerning is like playing the notes written in the outlines. Nothing else. No composition here. Just execution.
I'm a decent cook too. But I don't think that nutrition would require Ferran Adrià. My “cuisine” mot is: a cook only needs gram scales (possibly digital).

To Karsten:

we have written at the same time. But you better. Totally.

To Carl:

about the white space I copy myself and what I've written in a private mail:

I can say that iKern produces good results from 2005-2006. But I'm still working on it nearly every day and even now not only I have a to do list but I'm going forward. As a side effect of me being an engineer iKern functions. But the most important thing is the model. As my logo cryptically try to express the very core aspect of fitting is not the letter but the white space or, as I tend to think, the void. From this point of view my research is just the description of the many wonderful things that happen in a void space.

Igino, I sent you a private email about a use I put your Fell Type ornaments to. Did you get it?

Also, when I wrote what I did above it was as an idealist. But obviously it may be possible to do a great deal with a formula. I look forward to trying what you have made, and I thank you for it!

I agree with what you, Carl and Karsten have been saying. But I also think that saying that spacing is an art overstates things a bit while similarly saying it is simply a technique is to over simplify.

Igino: So this is still a work in progress?

Personally I love drawing type, but I hate kerning, which I am in the throes of doing thousands and thousands of right now. For me it is very exacting and excruciatingly boring at the same time. I don't see why it can't be largely automated, with some variables to be set by the person using the program.

The next time I do this, I hope I don't have to.

What I meant is, why smooth out the spacing with 21st century techniques, but leave other characteristics of the typeface in their original form?

The typeface was originally designed to not be kerned.
~

For instance, compare (what I assume to be) the original spacing, above, with the new spacing, below.

Isn't the shape of the W, with its angle of slant greater than the norm of the typeface, designed to address the spacing criteria of the available technology? By kerning after the W and T, a cramped effect is created, at odds with the letterforms.

You make a good point here Nick.

On the other hand, do you think the 1st spacing instance was really ideal even given the forms*?

*Which I agree were designed for their context.

I don’t see why it can’t be largely automated

It seems to me like it depends on what kind of color/notan/balance you get - and perfectionistic you are. If you make those things really integral to your idea/design I think you can probably end up keeping you hands dirty because the shapes and white space have to be designed together - no? But I suppose it does come down to if you approve of the results, and if you have to fortitude to fix what might be just okay. It might be that having "a leg up" let's you finish that race. Or that it enables a designer to walk away from an undone job too easily satisfied. In either event it is not the tool to blame or to be praised.

Yes, it is critically important to design glyphs in context to space well. But once they space well there is still kerning to do. At the very least, letters like WVTY in many combinations benefit from kerning, as well as punctuation. And adjustment of many others will result in more even color, no matter how well designed they are.

But I agree with Igino that the ideal kerning is so extensive, delicate, and boring that a good program is likely to save you from blunders. Certainly, one should proof and adjust. But hopefully the program would have enough variables that are adjustable that you could also some batch adjustments, as well as individual ones.

I believe Karsten mentioned somewhere that Lucas de Groot did kerning of triples somehow, which is actually a benefit also, but the whole thing gets quickly out of hand. I know Lucas also has lists of kerning pairs for every language, to reduce the burden.

Any tool that helps reduce the drudgery of this, and make it accurate, I would certainly praise.

there's a difference between mechanically centered and optically centered.

You could in theory, teach a machine to behave according the quirks of the human eye, and I imagine you could teach a machine to kern how we like things to look optically.

But, at the same time, you're teaching a machine to create art, not demoting art to simply a practice.

Igino,

that's amazing.

"What I’m saying talking about iKern is that I think to have found a mathematical model that describes those “underlying rules of fitting”."

Would you mind describing the algorithm somewhat more in depth, given you don't feel it violates your interests by giving away trade secrets?

I do believe you can give us more info on the mathematical background without losing out (I'm not asking for the complete code, not even for the exact algorithm :-) ).

Actually, do you use sequences of several letters, "words" when looking at the optimal whitespace, or just pairs, triplets of letters? Do you use Fourier-techniques to find the "rythm"?

Szabolcs

Just wondering, how many of the type designers in the audience run their fonts through Fontlab’s auto-spacing process before proceeding to space by eye?

Never have, never will.
Hand-crafted, dude!
However, I have always made considerable use of class kerning.

>>> kerning of triples

Wasn't David Kindersley doing some work on kerning triples? I recall hearing him talk about that about 1976. & did he not write about it somewhere?

I wish I could be more specific, but it is all recollection at this remove.

powers

To William:

> Certainly, one should proof and adjust.

To Nick:

Without kerning I would obtain the third line in the example. Honestly I dislike it and that's why I kerned the font.
You pointed out that the fitting of the “We” couple in Western is too tight. Probably. And I explain why. The first parameter iKern uses is “width”. It's defined for four cases: small vs small (s/s), Capital vs Small (C/s), Capital vs Capital (C/C) and Not Letter. At the time I compiled the Fell Types I was theoretically convincted that a “C/s couple” should have the same width of a “s/s couple”. And I set the input accordingly. Today I've changed my mind and I think that a good starting point for a “C/s width” could be the average of the “s/s width” and the “C/C width” (like I did in the examples in my site). But I say more: I can understand that in the present case with Capitals and smalls so different in weight could be accettable to set "C/s width" = "C/C width2 to let the Capitals more detached (like they where in the metal era). The first line in the example is from the font as is now. The second line follows the last indication. I've made this lesson to show, in practice, that:
iKern uses (needs) the human factor;
the effort of definition is minimum: there is the possibility to make global choices that are inherited through the font (it's class kerning at the maximum extent);
a feedback can be part (essential) of the workflow. For this please follow the thread where Jos Buivenga describes a little our mutual positive experience working on Anivers. http://www.typophile.com/node/46268

> However, I have always made considerable use of class kerning.
But the exceptions ...

To Eben:

> I agree with what you, Carl and Karsten have been saying. > But I also think that saying that
> spacing is an art overstates things a bit while similarly > saying it is simply a technique is to over
> simplify.

I can be provocative. But from my point of view the discriminant is the complexity (meaning richness) of the model.

To Szabolcs

> Actually, do you use sequences of several >letters, “words” when looking at the optimal
> whitespace, or just pairs, triplets of letters?

I can type a word to see the output, I can use text files for a sort of slide show, I can load in memory the modified font and use it in any other application.

Igino, you may be able to refine the tool, and make it more responsive to the discretion of the user, but you haven't addressed my fundamental critique of applying automated spacing to pre-designed types, especially those that are historic.

The shape of letter forms is intricately related to the media for which they are designed, and its quality of spacing.
In fact, the shape of the letters and the space between them are part of the same gestalt, notan, or whatever you call it, at any point in time. The design process is not a unidirectional "(1) glyph, (2) sidebearings, (3) kerning".

Therefore, to apply 21st century spacing to facsimile 17th century type forms is inappropriate.
Of course, the unspaced Fell italic all-cap setting looks wrong, but it would have looked wrong back in the day, too, and would have been generously letterspaced then, and with discretion, wouldn't it?

Nick, I've understood your point of view but it's a bit too radical to me. Let's suppose I had the possibility to copy the original spacing from the original punches and realized a font this way. I would have been in error even in this case because it would turned out to be a digital font to be laser printed or inkjet printed without the feeling of the real spreading ink, the pressure, the right paper, the words, the reader ... Maybe I extend too much but I should think you think that the idea of a facsimile is unaccetable at first. Anyway I made them to use them myself and not for a culural operation. That why I think I'm freed from the boundaries of fitting. And for what I know they've been effectively used by others in modern contexts too. I sometimes use the ornaments in my professional documents.

[But the sad truth is that during last Christmas holidays I've tried to refinish the Fell Types to clean shapes. But AFTER HAVING SPENT FOUR HOURS for a single letter I stopped. Destroyed. It seemed to me a prison prospective. So the fonts will remain rough].

Just wondering, how many of the type designers in the audience run their fonts through Fontlab’s auto-spacing process before proceeding to space by eye?

You've got to be kidding, right?

So, Igino, how much does the iKern service cost? Does it output kerning as a .fea? Does it group similar shapes (aáäàetc) and provide execptions (Ta Tä)?

—K

Hello Nick, I think your question is good: Should a revival pay tribute to the technology for which it was originally designed, and thus mimick flaws or pecularities resulting from this technology? But this question belongs to one realm (revival ethics), and the question which asks if iKern does a good job belongs to quite a different one (spacing models). It would be odd to criticise iKern because it was applied to the wrong kind of typeface.

I feel reminded of a previous discussions about the goods and bads of DecoType/WinSoft's InDesign plug-in Tasmeem, a layout engine for typesetting Arabic which uses a special font format. Thomas Milo usually presents its capabilities with help of a high-end typeface of a special style. (By the analysis of this style he arrived at the technology, but in the end the technology is suited for any other style too.) To my amusement or regret (depends on my mood) this provoked some critics to blame him for propagating this particular style which they consider outdated -- and disregard the layout engine because of the style of the typeface used for presentations! A layout engine of course is "mere" technology and is ignorant of anything like designers' sentiments.
Same is true for iKern. If you object to a Fell being kerned at all, then I had preferred if there were no "automated" in "applying automated spacing to pre-designed types".  ;-)

It would be odd to criticise iKern because it was applied to the wrong kind of typeface.

Yes, quite right.
But if the use of Ikern on pre-designed faces is questionable, would it not be better to consider its merits, as a 21st century design tool, in the context of an original 21st century type design? Was there mention of it being used in a face by Jos?

> So, Igino, how much does the iKern service cost?
You can email me privately.

>Does it output kerning as a .fea? Does it group similar >shapes (aáäàetc) and provide execptions (Ta Tä)?
Certainly. In my site there are some examples with the features file too.

>Was there mention of it being used in a face by Jos?
Yes. http://www.typophile.com/node/46268

I just want to reiterate that I appreciate the your new Fell Types release— I hope you don’t find this ‘lively’ discussion repellent.

Mikey :-)

James, FontLab's autokern is not regarded even by them as very good, and I don't know of anyone who uses it. I haven't tried it.

Miguel de Sousa wrote here in Typophile that he was quite impressed with DTL's Kernmaster, as a first pass to save work. I haven't tried that either.

I am using Matrix Machine, but that is just supposed to be good as far as ease of use and class kerning. All the decisions are made by the designer.

Jos is the first designer I've heard of using Igino's service, and he is evidently happy with it. I am uneasy about it being a service rather than a program that is under the control of the designer.

The real scandal is that digital typography never even dared free itself from merely imitating the one-letter-per-block paradigm.

Hm, Gutenberg did use around 150 special ligatures (multiple letter combinations!) to emulate te contemporary look of manuscripts. So there goes your argument, k.l.
I guess that practice was thrown out when efficiency and mechanisation took precedence (Monotype, Linotype — matrices).

. . .
Bert Vanderveen BNO

To William:

I am uneasy about it being a service rather than a program that is under the control of the designer.

I've tried to explain the control that can be obtained from the designer side too. In the end the choices to be made are not that much has my lack of details can have suggested to somebody. I say that the vast majority of the parameters iKern internally uses is extrapolated from the glyphs' shapes that's why I talk about a "Tecnique". It's true that if someone wants to override the "glyph's nature" he can but I found it rarely opportune myself to be honest.
Anyway for what I've understood it seems to me that "services" are always existed and exist today. It seems to me that is easier to control a machine than a man.

k.l. -- The real scandal is that digital typography never even dared free itself from merely imitating the one-letter-per-block paradigm.

Bert Vanderveen -- Hm, Gutenberg did use around 150 special ligatures (multiple letter combinations!) to emulate te contemporary look of manuscripts. So there goes your argument, k.l.

Very good. Actually your example supports my argument:
Ligatures were nothing but a workaround to deal with what I consider the core idea and limitation of the Gutenberg paradigm: that each letter is regarded as a unit, cast on its own block. (For illustration, contrast this with writing which by its very nature is ligated and "fluid" as regards black as well as white -- there are no ligatures because everything is potentially ligated.)
What Gutenberg did was trying to emulate the image of writing, so himself he wasn't aware of what I call Gutenberg paradigm. That Gutenberg's numerous ligatures and abbreviations were finally abandoned from the case however is just a logical consequence, and clarified what this paradigm is about. In so far, from a strict Gutenberg paradigm point of view, ligatures are a reminiscence to writing, a very artificial one, alien to this paradigm.

I am sorry that I was and am not able to make my point more clear. For me it is evident but I admit it took years to arrive there. (And from there, current font technologies are just a subset of what they could be.) As long as you think of type design as filling your preferred font editor's glyph slots with nice outlines, some of them representing single letters, some of them ligatures, some of them diacritic letters, etc, and as long as this works well for you, there is no need to try or adopt an alternative paradigm.

But this now is really off the topic which was iKern and the Fell type.

Karsten Luecke, KLTF

Tracking

ChrisL

Correction of typo: In my above post it should read "Metrics Machine" not "Matrix Machine".

Very engaging thread. I note that some assumptions about handset type may be incorrect however.

There are many stellar examples of composition and spacing from the past, right from the beginning, Schoeffer's work on the Gutenberg Bible is an amazing tour de force.

The transition from Latin based text is largely responsible for the abandonment of ligatures and abbreviations. I worked with a special Latin version of Carter's Miller Text, that included all the characters and ligatures that would have been used in the Aldine work, and it was like being in composition heaven.

Though many historical examples of commercial printing are quite bad, particularly in the 19th century, and that may be the perceived norm, one does need to remember that there were no compositional strategies in place until Tschichold's standardization of typesetting practices.

Metal foundry type in the second quarter of the 20th century (and into the 1960s), especially that produced by the more renown European foundries, was extremely well thought out in the settings. And many foundry fonts did come with alternative kerned capitals, such as a W that did fit tight to the following lowercase letter. Or an A and V that could be adjusted.

Kerning was practiced for some long time. Moxon, in 1683, describes a device for shaving letters to fit. Many high-end contemporary letterpress shops routinely kern their type with mortisers and/or glider saws.

Gerald

Karsten said: is just a logical consequence

I am not sure this is right. It was a practical consequence to be sure. But where does the logic come in apart from that?

And ...ligatures are a reminiscence to writing, a very artificial one, alien to this paradigm.

This makes compelling reading but again I am not sure. At a certain point in history it seems to me ligatures become very typographic in their form. And their justification for existence stops being a harkening back to writing so much as a harkening back to previous type. So while I don't disagree on some level - this kind of statement strikes me as maybe too absolute.

And As long as you think of type design as filling your preferred font editor’s glyph slots with nice outlines

Here I am with you all the way.

BTW don't think that this is completely off topic because the existence of iKern in many ways provokes questions about what Kerning or letterspacing is about/for and how it relates to letter design ( if it aught to be integral or not ) etc.

Igino, I want to ask you about the scheme that iKern uses in more detail but I need to re-read what you have written to better frame my questions. Thanks again.

Hello Eben, indeed I mean logical consequence, and as such it is absolute. The Gutenberg principle means one-letter-per-block placed next to each other, and has some implications of economy, e.g.:
-- as few blocks as possible  [ligatures to be shaved off with Ockham's razor],
-- avoid kerning where possible  [because spacing of letters is, in this principle, defined by left and right side bearings; nicely expressed by Erik Spiekermann],
-- space all letters by about the same amount  [since each letter/block only has one left and one right side (bearing); an effect of this is that uppercase letters usually are spaced too narrow, with A L T V W Y causing gaps, can be seen in about any font at hand].
During one of the next conferences, have a chat about Arabic and ligatures with Thomas Milo.  :)

Ockham's razor!!! You have got to be kidding here. You are freely mixing "philosophics" with "mechanics". The Gutenberg paradigm? Do you really think J.G. worked out the letters*/spacing of B42? Based on what? B42 is inscrutable. No one knows for sure. There is speculation the letterforms are composites, where the hell does that put your one letter per block?

*type didn't exist, as a term, until 1713.

Gerald

As a further addition to my previous post. You are assuming the existence of spacing material. Metal pieces used between words or letters. There is no proof of this in very early printed material. We simply do not know. It is assumed in some circles that B42 had none. Which means letterform "blocks" likely included the spacing. Hey, movable type.

If you are looking for standardization upon which to base you model (which I don't disagree with—just your terminology/historical placements), you have to look not at early Incunabula but rather with the Aldine, which is where the mechanization of letterforms clearly turns on its head and, agreeing with Eben, the influence of the organically prescribed letterform wanes.

Gerald

When Carl wrote "To put an even finer point on it, many discussions have arisen here about the role of “white space”, which is exactly what Inigo’s program is concerned with. Evading the crucial lessons of spacing sounds initially like a terrible loss in a field where visual discernment has always been crucial. " I think he hit the nail on the head as he often does.

This is the question that a user of Igino's sytem has to consider. My interpretation of what he has said so far is something like "you can use my system to skip ahead and make refinements afterwards if they are neeeded". This is a tempting idea. But again, if you believe that kerning is not a chore to be dispensed with but a core aspect of type design it is still hard to think of iKern as something to bring into an ideal process. On the other hand, not all typefaces or process' need to be ideal I will admit! Nevertheless, the question remains.

Ockham’s razor seems like an odd tool to consider in concert with Gutenburg model of printing which had so very many special versions of letters and so very many ligatures! But this question of fitness of a reference aside, the thing that strikes me is that while your idea may be sound in some respects - that a robust design capable of fitting well with a minimum of fuss has deep utility, ( correct? ); the idea that this is the logical absolute and inevitable model we must consider is something that I ( of course ) am pretty skeptical of. I can sort of see the idea that the Gutenberg process might have an inbuilt tendency towards simplicity over time. That idea is a I think far from being either proved or disproved. In it's favor would be the idea that printed books must have taken on their own validity outside of the context of the written book over time and that once that context was established that the power of economic need would dictate a move to simpler and faster methods.

Part of the reason I object is that I think this idea is what Americans would call "monday morning quarterbacking", meaning coming up with explanations and justifications after the fact. It's quality of being "Ex post facto" might lead you to the word "absolute" in the sense that this model does indeed dominate at the moment. But will that status last? Does it have to? Is it actually best? I am not at all sure about any of these things.

When I read Morrison on type design and on Incanumbula and especially on Jenson I get that same feeling. Like I am not reading an incisive theory that explains something in a deep manner but rather that an explanation or story that has the lone virtue of fitting the existing data is being conjured up. Get a bit more data and that story could be tipped over...

Reading Van Krimpen give me again, the same feeling.

Similarly, if iKern is based on fitting certain kinds of forms and has assumptions built into it about what letterforms are probably like ( the story) rather than having a way of sensing them directly, like the Kindersley system did, then it is not going to be a very robust kerning tool.

Unfortunately I still don't quite understand the mechanism of iKern.

Maybe I will eat my words. But that's how it looks to me now.

I don't know about the rest of you, but I view the spacing and kerning of a font to be part of the design. Muling over the spacing gives me an insight into how it all fits, and more importantly where it does not fit. Countless times I have redrawn glyphs because of some spacing issue. So I don't understand the approach of drawing the glyphs and then applying the spacing.

Drawing and spacing are one action, not two.

I also think that given the class kerning tools available in FontLab kerning is not such a tedious process, generally (there are those exceptions!)

JamesM

I do agree that spacing and kerning are part of the design. James has an important point about it often being necessary to redraw glyphs to take fit into consideration. However, once the basic design is set, spacing and kerning still remains a repetitive chore that can benefit from being systematized and automated.

I don't have a detailed knowledge about how iKern works, but if it's anything like my fantasy spacing/kerning tool, it will allow me to control several parameters related to spacing and kerning with something like sliders, and have those parameters applied systematically throughout the font.

I'm really curious about the points Eben raised though about how much iKern relies on assumptions about letterforms. Has it been used to space and kern non-Latin fonts, for example?

Mr Bieler, I think there was a fundamental misunderstanding.

In my previous post I wrote "[Gutenberg] wasn't aware of what I call Gutenberg paradigm" which says:
(1) "Gutenberg paradigm" is my conception/abstraction, not Gutenberg's.
(2) "Paradigm" should indicate that I am speaking from a logical/philosophical point of view, not from a historical point of view.

Again, I am not attempting to give a historical account of what Gutenberg did or thought, but am talking about a paradigm (as the sum total of principles or ideas or conceptions or however you call them) that current digital font formats have adopted from an earlier technology.
(Eben, by their nature, paradigm, principles, ideas, conceptions are not tangible objects that I could "prove" by simply pointing at them.)

An even earlier post's "Gutenberg’s one-letter-per-block" was indeed a lapsus.

* * *

Re: Which means letterform "blocks" likely included the spacing.

This is what I wrote in an earlier post, "that each letter is regarded as a unit, cast on its own block" and in the previous post referred to this as "one-letter-per-block", abbreviated as "letter/block". We don't disagree. And there was no mention of spacing material in any of my posts.

What I would be hoping for in an extended auto-assist in kerning such as Ikern would be an extension in what we now have in class kerning. It would have identify further categories, under the control of the designer. That way the drudgery would be reduced, while leaving decisions in the hands of the designer.

My first question to Igino was if I could license iKern, but because that wasn't possible, I decided to have a go with Anivers to see how iKern could stand out, because I did see the app perform very good on the Fell Types (apart from the discussion on *how* this should be done). Igino did some tests based on an earlier release of Anivers and was able to capture the look and feel of that release with spacing beautifully and consistently translated to all glyphs (family wide).

Everything should be under control by the designer, I'll be the first to agree with that, but while I didn't push the buttons and let my laptop crunch the calculations for two hours, I never had the feeling I wasn't in control. The same goes for the whole kerning process. I did request for an early (spacing) test run on the other styles and weight so I could change certain glyphs if needed.

I regard the whole process and collaboration with Igino as a joy and that is why I asked him if I could mention this, because I thought this time saver could benefit more typophiles.

To see a result of this project you can download Anivers Regular *here* for free @ MyFonts.com.

Why is there kerning between W and T, but not between L and A?

Maybe Igino can explain the technicalities of that when he chimes in. I did numerous tests and ran more than twice through every kerning pair. It looks good to me. What do you think needs adjusting?

What William and James are saying makes good sense to to me generally. But I think the ultimate kern/side bearings assist would either

• offer the designer a series of combinations to be spaced and then take those responses and create the side bearings and kerns needed to achieve the specified optical result. I recall somebody wrote a really lovely request for fontlab that described something like this.

or

• Examine the letterforms in being drawn a quasi-optical manner and suggest common pairs and a possible spacing given variables like optical size which the designer would then override on a case by case basis. Again letting the designer design, and sorting out the most efficient ( least kerns and least extreme kerns) code based on those design decisions.

Karsten wrote: ...by their nature, paradigm, principles, ideas, conceptions are not tangible objects that I could “prove” by simply pointing at them. And he is of course correct. "Prove" is the wrong word.

But instead maybe I can say something like "make a case for". I think if you are going to use words like absolute then you have to build a strong case at least. Also, I may not be getting the thrust of your argument. Were you saying that a Gutenberg paradigm was one that leads "logically" and "absolutely" (I might say inevitably), to simpler shapes? If that is right, then that's a pretty strong statement worthy of some examination and critique, and needing, in turn, support and reasoning before it could be accepted.

BTW From what I can tell ( which is less than I would like) Gutenberg's special sorts existed more to allow perfect spacing of the line than as optically driven variants for the fit of letters except when they modeled a ligature which may or may not have existed for better fit.

Probably because you can't plug in just any ole typeface and expect it to magically work.

The Gutenberg paradigm is brilliant and how can one not love the line ligatures were shaved off with Occam's razor but I think this may be a case of the cart before the horse.

The common liturgical hand of Gutenberg's period WAS mathematically simple. The textura has three basic strokes. Letterform fitting to the mechanical bounding block with such a typeface is, relatively, an easy process to work out. In fact, this simplicity is considered, in some circles, the rationale that allowed for mechanical writing to have developed.

With the evolution to roman based faces during the period of Incunabula, the process became much more complex as the letterforms were more complex, and the mathematics of fitting were more complex.

The transition from Latin to the vernacular likely plays a greater role in the disappearance of the multiplicity of sorts (ligatures, abbreviations) than a movement toward mechanical simplicity.

That the printing process was early on considered mechanical writing lends credence to Eben's "the idea that printed books must have taken on their own validity outside of the context of the written book over time and that once that context was established that the power of economic need would dictate a move to simpler and faster methods." Finding the phase where the mechanical emulated the mechanical isn't all that historically precise but it would be at this point rather than earlier development where the concerns here become more valid. The period of Incunabula is separated out uniquely in printing history for a number of very good reasons, it is an arbitrary distinction, but it helps to look at the development of printing as having a birthing stage, larvae, that is unique from all that follows. When the pupae emerged full grown is where these considerations are best developed, not prior.

So I don’t understand the approach of drawing the glyphs and then applying the spacing.

This begets a technique question I often wonder about: how many type designers are kerning while they design as opposed to finishing the drawing and spacing first?

To Eben:

Similarly, if iKern is based on fitting certain kinds of forms and has assumptions built into it about what letterforms are probably like ( the story) rather than having a way of sensing them directly, like the Kindersley system did, then it is not going to be a very robust kerning tool

Unfortunately I still don’t quite understand the mechanism of iKern.

To Terminaldesign:

I don’t know about the rest of you, but I view the spacing and kerning of a font to be part of the design. Muling over the spacing gives me an insight into how it all fits, and more importantly where it does not fit. Countless times I have redrawn glyphs because of some spacing issue. So I don’t understand the approach of drawing the glyphs and then applying the spacing.

I've never received a font without a working spacing (why can't it be refined?).

To Jongseong:

I’m really curious about the points Eben raised though about how much iKern relies on assumptions about letterforms. Has it been used to space and kern non-Latin fonts, for example?
Ikern has no assumption about letterforms. Any geometric shape can be handled (look at the V and A of the Italic Fell Types and how they're handled elsewhere).

To Nick:

In the image above you can see the word LAWT a) without kerning; b) with the output iKern produces before cutting the values under a certain threshold (in this case 3 so the kerning between LA will disappear); 3)the actual version.

Both L and T are detected by iKern as letter where the “space” around the outlines is important for the “identity” of the letter. I use the term “bodiliness”. Internally is a property referred to the left or the right side of a glyph and is a value between 0 (no bodiliness) and 1 (maximum bodiliness). L and T usually have 1. O usually has 0. These values are calculated by iKern. After all calculations, during a post processing phase, there are some further actions on the kerning values depending on the bodiliness values of the glyphs involved. I won't go into details but there is, among others, an action to set a minimun distance between the glyphs proportional to the maximum bodiliness values. In the case of Anivers this distance was 36 points (it can be manually tweaked if one wants. I always let the default value). If you measure, in the unkerned condition, the minimum distance between the LA is 34 points. The measure between WT is 31 points. That's why the LA kerning becomes (36-34) = 2 and the WT kerning becomes (36 – 31) = 5. As I said before in a post-post processing phase the kerning values under a certain threshold are cur away (the value is choosen by the designer).

What do you think needs adjusting?

Well, if it were my face, I might put more space between the L and A, but then again, I might not.
I don't believe in absolutes, and that includes any "underlying rule of fitting".

I once rekerned Palatino (in Quark XPress) for a series of publications I designed in 1991. I wouldn't say my spacing was better than the default, but it was something I decided would work for my design, which involved condensing the face around 90% horizontally (the horror!) I've dug up one of those pieces--text at 10pt-- and scanned it, and reset it with the default spacing for comparison:

a minimun distance between the glyphs proportional to the maximum bodiliness values.

Surely the bodiliness value that's relevant to L_A and T_W is for the negative area immediately adjacent to the horizontal stems, and with that stem height--not for the glyph as a whole.

Well, if it were my face, I might put more space between the L and A, but then again, I might not.

Different designers make different dicisions.
And often have different oppinions ;-)

I took a look at your latest font Paradigm to see how you did "LAWT". I myself would kern AW less. Especially when the face is also intended for smaller point sizes.

Paradigm isn't intended for any particular point size.

I spaced the capitals anticipating that it will be used in the kind of setting where typographers will add letterspacing (see below), or if using an all-capitals setting in text with zero tracking, would choose small capitals, where I have built in a little more space to the A_W combination:

You can play around with the tracking here.

Eben -- this kind of statement strikes me as maybe too absolute
Karsten -- indeed I mean logical consequence, and as such it is absolute.
Eben -- I think if you are going to use words like absolute then you have to build a strong case at least.

But Eben, it wasn't me who introduced this term! I admit it was my fault to not having ignored it, or at least putting quotes around it to indicate it's not mine.

Were you saying that a Gutenberg paradigm was one that leads "logically" and "absolutely" (I might say inevitably), to simpler shapes?

I did not say it "leads ... to simpler shapes" -- didn't speak of simpler shapes at all (this discussion is about spacing), rather said that the paradigm as such implies a certain aspect of economy (but then, it was another fault to respond to the ligatures theme which is secondary in a discussion about spacing).
I did not say it "leads 'logically' or 'absolutely' ... to simpler shapes" -- and wouldn't because of an aversion against Hegelian thinking, rather I tend towards a Kantian/Peircean understanding of "logic": If you think of it this way, then you automatically imply that. The "if-then-imply" is what qualifies as "logical".

This thread is like an Indian carpet with comments weaving into each other making a colorful topic.

@ Nick :: Thanks for the explanation. I understand very well why you did Paradigm the way you did. I choose fairly loose kerning for Anivers because I thought that fitted my face better. iKern can do both of course.

To Nick:

Surely the bodiliness value that’s relevant to L_A and T_W is for the negative area immediately adjacent to the horizontal stems, and with that stem height—not for the glyph as a whole.

I haven't said how "my" bodiliness is calculated and how its treatment is. As I already said, I can't go into details so I'm trying to express myself in the easier way possible. The fact that, as default, I've put the minimum distance as minimum as possible is functional, in my opinion, to preserve, as much as possible, the rhythm. For the same reason the AV kerning is set as wide as possible. LA and WT have a lot of "air"? Let's minimize it; AW has little "air"? Let's maximize it.
Anyway it's not important. I'm glad of your comment. What have you done? You have isolated a problem having in mind a solution, thanks to your experience. Imagine now to try to formulate the problem and the solution in mathematical terms, finding general concepts to put into a type oriented framework. Imagine doing this many many times until you feel the framework is grown and is "robust" enough (in this case only you could know!). This way you could give iKern the right to existence!
[Before you reply: I don't think you don't give it. I appreciate your positions because are honest.]

I'm not trying to convince anybody to change religion. And, listen well, I'm not just promoting myself as a service owner. All I want is that people begin to consider some degree of automation not being so bad even in terms of quality. And I want that people know that there is the chance to adopt a different and productive workflow.
The way? Try, use your eyes and judge yourself if it's worth.

@ Nick :: Thanks for the explanation. I understand very well why you did Paradigm the way you did. I choose fairly loose kerning for Anivers because I thought that fitted my face better. iKern can do both of course.

Hi Jos, you surely know what I'm trying to say better than anyone here!

Strong sense of déjà vu...

http://typophile.com/node/10317

Eben, what's that "Kindersley system" you mention?

Why I'm sceptical about Igino's service is not so much because of some kind of dismission of automation per se, much more its the secrecy igino makes about the actual method employed which makes me wary. In the case of cars, e.g. I approve of ABS, because -- while I don't have the exact paramters of the IC-chips and the exact parameters of the functions employed or the exact differential equations set up, there is publicly available information about how it works, its basic principles in details, even which equations are to be solved by the system. So I can counter-check and trust my confidence into it. I couldn't do that in the case of iKern.
Come on! I even know the basic principles behind nuclear power! I don't care about the actual measure how much they insert the graphite bars to slow ("moderate") the reaction, whether its 1cm or 10cm, I'm sure they know how much they have to (they don't want to die after all either), but I know what they do in principle. And for knowing that there is still no threat that I will construct a nuclear power plant myself and pose a business competitor.

While I said I do not dismiss automation, as division of labour and employing new tools, as well as automation is the way to become more efficient, it's the way of progress (though note that the German word for progress: Fortschritt etymologically does not mean a change toward better, not even forward, as the latin word suggests, but rather only away from the status quo... good to keep in mind); however, there is a common understanding that automation always goes hand in hand with lower quality without exception. Or why else is always "hand-crafted" seen as a label of quality? Automation is viewn completely without drawbacks or loss of quality only, if it makes something possible that was unachievable before, not however, if it makes something just more effective or efficient, faster to deliver.

Szabolcs

Hi Jos, you surely know what I’m trying to say better than anyone here!

I guess the proof of the pudding is in the eating.

Karsten, Thanks for making that clear.

Nick, Quite right.

Szabolcs, you can read about the Kindersley spacing model here. Click on the link for "spacing".

http://www.kindersleyworkshop.co.uk/

He was involved in making a kind of electronic eye with some to judge things. It wasn't perfect and some letters confused it. But mostly it worked fairly well. Hermann Zapf was keen on it.

James P asked: "This begets a technique question I often wonder about: how many type designers are kerning while they design as opposed to finishing the drawing and spacing first?" I think that you will find that 100% of the best type designers space and create their glyphs back and forth a great deal beacause the "white" is part of the design.

Why I’m sceptical about Igino’s service is not so much because of some kind of dismission of automation per se, much more its the secrecy igino makes about the actual method employed which makes me wary.

Ikern is actually intended to make fonts and not to talk about making fonts (ironically this attitude shares very much with the “done by hand” practice). They way I answered to Nick, the only one, about a technical and not theoretical question, I use to do in private mails with people interested. And if somebody would have asked something the same way I would have answered just the same way. And after a short debate the landscape would have been clearer and, I know myself, I would have said more than I should have. By the way the general principles of iKern are written on the only one page of my site: "Proximity" "Bodiliness" and "Rhythm". Has anybody asked something about them? If nobody asks I must think people alreay knows or the interest is somewhere else.

By the way, by all my abovementioned (and explained) scepticism: I find the example in the old thread linked in very impressive! And I feel honest appreciation for Igino's work! (But the bitter taste remains, and, maybe you not, but I'm a person who dislikes employing a technology I do not understand sufficiently, even if it "works").

Szabolcs

This whole debate is very interesting, but it's mostly theoretical. I would like to know more how iKern actually works – the theory behind. But I understand if you will not answer exactly, because of all the work you put it in iKern.

I read the definitions on your site and also another typophile thread with a little more info. I think I can understand “bodiliness”, but not “proximity” or “rhythm”. And a fact that iKern can use overlapping value from 0% to 100% confuses me even more.

I am sure that internally the representation of the character is in raster image, because it would be too difficult to mathematically compute parameters directly from bezier data. Is that so?

Did you discover the algorithm of iKern all by yourself or did you help (or get inspiration) with previous attempts with autospacing?

Szabolcs -- but I’m a person who dislikes employing a technology I do not understand sufficiently, even if it "works"

I share this attitude. Yet, do you [need to] know how exactly InDesign's "optical" kerning or KernMaster work? Would you ask Adobe to publish InDesign source-code so you can study it and decide if you like how it works?  :D

Eben: But again, if you believe that kerning is not a chore to be dispensed with but a core aspect of type design it is still hard to think of iKern as something to bring into an ideal process.

I think it is a mistake to say that kerning is 'a core aspect of type design'; rather, spacing is a core aspect of typography. There are two important points made in that statement.

The first is that kerning is a particular mechanism of spacing, and one which presumes other mechanisms, such as sidebearings. Further, it is a mechanism that presumes that spacing is something that is achieved in two (or potentially more) additive passes: default spacing and then adjustment of specific sequences. Or, to use Einstein's terminology, general spacing and special spacing. Since it is a mechanism, I don't think you can say it is a core aspect of type design; it is, rather, an important step within a particular kind of font manufacture. This becomes obvious when you consider that exactly the same spacing could result from a different mechanism.

The second point is that spacing is really an aspect of typography, which was more generally recognised in typesetting prior to the desktop publishing revolution, and is reflected in manuals of that period. [By the way, Eben, Kindersley's spacing experiments need to be viewed in the context of typography (and sign making, his initial inspiration), not type design or font manufacture. In many respects, Kindersley's work -- about which very little solid material is available; I have copies of all the published stuff -- is closer to the kind of thing that Karsten is talking about that conventional font spacing.] The conventional model of general and special spacing via sidebearings and kerning data is, at best, appropriate for a particular size of set type (ignoring other factors that may recommend adjustments to such spacing). As such, this font spacing should be seen as a kind of helping hand, and not as a replacement for typographic control, whether applied by an experienced typesetter or automated in some more or less sophisticated way relative to type size.

I think it is a mistake to say that kerning is ’a core aspect of type design’; rather, spacing is a core aspect of typography.

In my experience with iKern I would say:
spacing and kerning are just two faces of the same medal. I state so because, in my application, autospacing and autokerning are done by the same algorhythm (and same routine) sharing the same data and, as a consequence, I note that the statistical distribution of all the kerning values is a gaussian bell centered on the zero and with very low variance. Translated means that only a few (averagely 10,15%) of the all possible couples turns out to have a kerning whose absolute value is above a certain low threshold (default 3 points for 1000 UPM, 5 points for 2048 UPM. Can be choosen by user anyway). I also see this as direct application of the “first spacing, then kerning” adagio.

The second point is that spacing is really an aspect of typography, which was more generally recognised in typesetting prior to the desktop publishing revolution, and is reflected in manuals of that period. [...]. As such, this font spacing should be seen as a kind of helping hand, and not as a replacement for typographic control, whether applied by an experienced typesetter or automated in some more or less sophisticated way relative to type size.

Really true. I think the limitatations are in the font format and in the applications. Apart from choosing a size and tracking accordingly, what else can a typographer do? I think the possible “operative” parameters are far more. Examples:
Commenting Nick I talked about a minimum distance between glyphs iKern forces under certain conditions. First: this minimum distance can now be choosen by me or the designer but should have to be choosen by the final user. Second: if I vary the tracking the kerning should consequently change. But I can do it only with the iKern internal traking for the unique istance I produce. In applications kerning remains constant and I loose any possibility of control.
Another parameter iKern uses is “kerning strength”. I noted that text setting require a looser kerning than a display one. So generally I choose the looser direction because “it's better too loose than too tight”. And I have to choose because I can give away only one istance.
By design iKern let you choose one of the infinite possible spacings. But in having to choose between the theorical best and a “manufacturable” one you have to choose the second because, in the end, a digital font is a product that must work in most cases as possible.
Now I've moved most iKern parameters in a “post-processing” phase so I can tweak with them without recalculating everything. So I would say that, in my opinion, the iKern viewer and postprocessing panels are, together, the tool that now permits the wider range of typographical control on a type. Obviously using the iKern file format.
In the long run this is the direction I see for iKern. With the today service I'm only trying to make this concept acceptable.

John, I was too quick to use the word kerning. Yes I do mean spacing not kerning. You make some great points especially Re: using type at multiple sizes and the impact that has on the spacing that will be appropriate. And most of all the potential unreliability of the spacing you set out as a designer. And so I agree that it is probably wise to consider it as 'helping hand' in some sense. Also what you are saying certainly does add to the notion that a robust design, one that takes spacing degradation with less trouble, is very desirable, which is in a separate point.

Still, don't you design as Noordzij says "the white of the word" along with the black even if that white is in some cases going to be more provisional than you might like? Or do you design first and space afterwards?

It seems like people are increasing interested in having type's use specified and that designers are increasingly designing for specific contexts and this seems desirable. It seems to me too that if you are designing a complex family with versions for narrow pt size ranges like Arno pro the helping hand of spacing can be seen as somewhat less soft suggestion even if still is a "helping hand".

I would like to hear more about the distinction you are setting up between Kindersley’s ideas and Typography vs. Type design. I see that the two are different things but they seem a bit like a venn diagram to me with type design's question of spacing overlapping with the practice of spacing in Typography. But maybe I am missing something. Am I?

Yes, the Venn overlap.
Spacing is an intelligence that can be handled in different places.
It can be put in the layout application (InDesign Optical), or in the font (Metrics), or it can be manually applied by the typographer.

It should also be understood that such intelligence is discretionary.
To illustrate, consider this example.
First, this is how it sets default. There are negative sidebearings on v and y, which implement the idea that the white space between letters should be evened out.
However, there are at least two conflicting principles in spacing, the two main ones being "controlling area" versus "controlling proximity". Here, in the v_y combination, it seems to me that space has been overly privileged against proximity. I realize that this is my aesthetic, today. In the past, I would have been OK with the overlap. But not today. So:

1. Default.
2. With added kerning to provide separation, the way I like it.
3. But if this is built into the font or application, it doesn't work when tracking is tightened for a display setting.
4. So, according to the "tight but not touching" aesthetic, yet further spacing has to be added, between e_a and v_y.

So even with a sans serif font, it may be necessary for the type designer to provide optical size variants, and even then, the layout application may be able to offer different "flavours" of kerning. And even then, the typographer may be able to say, screw that, I'm going to exercise my responsibility, turn on manual overide, and blow up the f****** Death Star.

The optical size issue definitely adds a layer of complexity to the issue, as several have written here.

Slimbach has argued that with larger optical sizes spacing of round to round letters is often are best changed differently from straight to straight letters. If this is correct, tracking won't do the trick for the best spacing at larger optical sizes. There is also the issue of adjusting kerns for more than one character, as Nick notes. This can get very complicated, but there's also the question of when it ceases to make any real difference.

I gather what Igino is arguing for is that his program should be built into type setting programs for the best result. And perhaps having more variables under control of the type setter as well? That is intriguing and makes sense to me. In a way Adobe's "optical spacing" is maybe something like that, but of more limited use than what Igino envisages.