Jacques --

At one point during the development of Whitman, I spent several hours with various dictionaries -- French, German, Spanish, Italian, Portuguese, Swedish, Danish, and Dutch -- to identify accent combinations in either the first or second positions. I was looking for possible cap/lc kerns. Not being at all fluent in most of these languages, I'm not sure which words might have been completely unlikely to show up as the first word in a sentence, but I got into an obsessive mode.

I was also interested in finding which accented characters apeared in initial or terminal positions, in which languages, to consider whether they required kerning with various quotation marks.

It was an interesting exercise. I can try to find my list somewhere for you, when I get a chance.

I believe John Hudson has developed some extensive kerning lists along these lines.

-- K.

Hi Kent,
sounds interesting. Maybe we should not only make a list of possible pairs, but have a list of words, which shows it as well. And of course with quotation marks or other signs.

---Jacques

Good kerning is based on linguistics. Which is stuff like this
http://www.typophile.com/cgibin/show.pl?30/13106
is important.

hhp

Now that I look at the Typophile kerning pair text file, two posts back, I notice a couple of probable typos. There's a "t." on the same line in Lower/Lower as "qu", suggesting that it wanted to be "q." instead (borne out by the fact that there's another "t." a couple of lines down), and for some reason in Cap/Cap, the letter "M" has become a kerning pair all by itself. "MC", maybe?

Dear Jacques,

This kerning files are very usefull, please test your fonts with them, and thanks to sudtipos to upload to all of us.

http://www.sudtipos.com.ar/test01.txt
http://www.sudtipos.com.ar/test02.txt
http://www.sudtipos.com.ar/test03.txt

M.

Hi all,
this is going well. Especially thanks to sudtipos.

I am busy making a own list. Will be here soon.

Keep the combinations coming.

---- J

>

I would suggest adding a second lowercase letter to your initial cap kerning samples. A "Ta" combination can look like anything, a "Tap" letter combination will go a long way towards helping you kern correctly.

Tap Ted Tch Tip T

I'll note that I find the whole notion of kerning *pairs* rather than *classes* almost terrifying in this era of ever-larger character sets. I'm glad I never have to work with pairs any more.

==

I'll echo Mr Phinney's sentiment regarding classes. Once I got my head around the concept of applying metrics by class in the FontLab workshop at TypeCon, I was blown away by its potential and ease of use. Bye, bye pairs. Don't let the door hit you in the butt on the way out.

BTW, on the ATypI list in June, Frank Blokland wrote:
"At DTL we use a basic kerning file for our databases that contains a
selection of roughly 5000 pairs that include combinations for Western
and Eastern European, Cyrillic, Greek, smallcaps, old style figures, etc.,
etc. in combination with KernMaster."

Now that version 1.08 has been released*, you can download the FM (demo) installer and access the file named Complete.krn from the FM folder. Or you can always just buy FontMaster! It has some truly unique features.

* http://www.fontmaster.nl

hhp

Hi all,
thanks for all your work this far. But be ware of sending lists of all possible kerning. The purpose of this thread is to create a list of all pairs which are really occur in the countries which can use latin 1 font*.

For example:

> Would you prefer to ....

It's useful to mount a two-fronted attack:
1) Use control strings.
2) Use real words. Two lists: one based on frequency (like "the"), the other based on how problematic it is (like "wyvern").

--

BTW, great effort! Thanks.

hhp

Kerning in Adobe OpenType Pro fonts contains many pairs which couldn't appear in a real written text. There is a limited amount of Czech accented letters, like ccaron, scaron or uring. These characters are used in Czech word only and cannot be combined with French / Italian etc. letters. So the combinations like Tcaron+acircumflex or Lslash+Yacute are absurd. I think it would be great to create linguistic-based kerining pairs. There could be several lists of pairs - for Western, Eastern European, Cyrillic and other encodings. (May be the problem is, that Adobe OpenType Pro fonts simply kerns letters T+a with all possible accents.)

>(May be the problem is, that Adobe OpenType Pro fonts simply kerns letters T+a with all possible accents.)

I believe that this is exactly why you find the kind of absurd combinations you cite. But, while this would be a disadvantage in a typical Type 1 font (because bloated kerning tables slow down processing), since Adobe OT fonts take full advantage of class-based kerning, these combinations take virtually no extra code (depending, of course, on how the classes are defined).

So, in the case of OT fonts, I think more might be better -- i.e., better to be thorough and have a few absurd combinations, than not to have accented pairs. Also, since it takes no real overhead, casting a wide net is probably a lot more efficient than trying to analyze every conceivable language and identify only the (currently) possible combinations.

In OT, that is -- with class-based kerning. Not in Type 1. Jacques's list will be useful for Type 1 kerning.

-- K.

Just to expand on Kent's comment, removing those "nonsense" pairs would in fact make the font larger, when using class kerning!

I'll explain how it works with only a little simplification. There is a matrix of left vs. right classes. Any exceptions are additional line items that must precede that matrix. If an exception is necessary to avoid a collision or because the kerning result was poor, then we'll try to do that. But suppressing "nonsense" combinations would simply add more exceptions, and bloat the font, without making it any better--it's not like they hurt anything.

Regards,

T

Kent and Phinney,
you're right. I didn't know about OTFs actually. I haven't been very deep into that chapter yet. Indeed, my original question is to be seen within the realm of type 1.

Please correct me if I'm wrong, but if one works with classes, like phinney explained, isn't there a chance of becoming colliding glyphs or exaggerated kerning? If so, you still need a list (of all pairs that make sense) to be used for controlling.

---jacques

Filip,

> I think it would be great to create linguistic-based kerining pairs.

This could be a solution, but it st more practical to have fonts, which includes all languages included in a western type 1 font. Otherwise you'll end up with maybe 5 or even 10 different fonts. This is doomed to go wrong. Imagine your font is being used by a multinational, operating in the US and in france.

--Jacques

>Please correct me if I'm wrong, but if one works with classes, like phinney explained, isn't there a chance of becoming colliding glyphs or exaggerated kerning?

I have not delved into class kerning yet, so I speak only form a theoeretical understanding, not practical experience. But the theory is that if you have equivalent sidebearings in your base letter and the rest of the members in the class, then what's good for the base is good for the rest. There are, of course, cases where this breaks down -- for instance the kern for Va may not work for V{agrave} (depending upon the design of your grave). That's when you'd create an exception, I suppose. The other option might be to define a second class of a-like glyphs which doesn't include some accented a's specifically for kerning with overhanging capitals -- that sort of thing. But, of course, too much of this -- either exceptions or extra classes -- starts to enlarge the font and also the work, almost defeating the purpose of using classes.

Class kerning is a powerful tool, but it requires planning and careful logic in setting up your classes, as well as designing your glyphs. It seems like it would work better with consistent, regular designs than with idiosyncratic, irregular designs.

-- K.

>

> > Class kerning is a powerful tool, but it requires planning and careful > logic in setting up your classes, as well as designing your glyphs. It > seems like it would work better with consistent, regular designs than > with idiosyncratic, irregular designs.

It also requires applications that consistently recognize class-based kerning. We're not quite there yet.

From the little I know about it, class kerning is sounding like the pantograph.

hhp

[email notification test- please disregard]

I have been pasting words & combinations/pairs into fontlab & I noticed that the accented characters are represented not with the typed chracter in the paste area but with a slash, the letter & then the accent name. Although all the resources I have found are great; I'd like to find (or build) something geared to Fontlab.

This would mean making a document where the pair 'T, a acute' would be written 'T/aacute'. I could do a search & replace on one of the larger docs like http://www.sudtipos.com.ar/test01.txt , but that seemed like a half solution.

It sounded like there had already been some work done making real examples to consider along with the pairs like,

more kerning resources:
Leslie Cabarga's Kern King
and if you're runing a Mac, OS 10.3, you are lucky enough to (someday) be able to use the AMAZING MetricsMachine. (I'm not sure if it's readily available yet, but contact Tal)
why did i buy this darn PC laptop???

As a starting point - this is a very basic list, mind you - there's a file on the download page here:

http://www.typophile.com/downloads/kerning_pairs_1.0.txt

I should think that Peter Bilak has two very useful looking tools on his site.

http://www.typotheque.com/type_utilities/index.html

[ This thread moved to "Design" ]