> I have the cleartype fonts...

how?

Christian,

I don't think Microsoft has officially released them, so those are probably leaked copies, i.e. warez. But Simon should know for sure.

Hi,

Pre-release versions of the fonts are included in the Windows Vista beta version distributed to many beta testers around the world. A few sites (through ignorance or mischief) zipped up the fonts and posted them for download. We’ve been quite successful at having some of these removed, and are working on the others. As anyone who’s had their stuff posted on the web will know it’s an uphill battle to get it taken down.

Christian, if you can forward me the site address (simonda@microsoft.com) we’ll get in touch and ask them to take the fonts down.

Oh, and the CT Collection fonts in the Windows Vista Beta have OpenType kerning - not old-style TTF kerning.

Many thanks, Si

What's this difference between old-style and OT kerning again?

And it seems strange that somebody would
end up distributing a handicapped version.

hhp

Hrant:

Some people simply don't know any better. I'd wager that these same individuals don't know what kerning is.

sii:

Thank you.

Most people might think this is a simple topic: cleartype fonts should have kerning. I think that's okay for print. But my understanding of CT *today" is that the rasterizer does not care much for horizontal hints, right!? Now for a brief detour: Most type designers I know are of the opinion that if *one* thing goes wrong with a text face, pffffft goeth the quality. Well, if you're not spacing meticulously (at screen resolutions), what good are unhinted kerning pairs added to said mix? Hmmmmm?

Simon said "Oh, and the CT Collection fonts in the Windows Vista Beta have OpenType kerning - not old-style TTF kerning."

What happens when I generate a OTF Font from Fontlab? Does it traslate any kerning data into OTF format? If not, what happens when you include kerning data in a CFF OTF font ( what Fontlab makes) and then add OTF kerning data? Are the applied only seperately? Or could you get them applied together resulting in a double kern? What environments apply OTF kerning - Just Vista/Longhorn? Does this mean you have to have different OTF for mac & windows? I thought we were meant to have a unified format here...

No, It's helpful. Thanks. But yes, I want to know more. I'll ask the fontlab folks. Do you know anything about CFF OTF?

So, the main differences between 'kern' table (old fashioned) kerning and GPOS kern feature kerning are:

1) the latter supports class-based kerning and the former does not
2) the OpenType spec says that OpenType CFF fonts only use the latter

FontLab will generate either or both types of kerning, for OpenType fonts with either TT or CFF outlines. The mysteries of when it generates which are determined by preference settings in FontLab. In FontLab Studio 5, it's under Tools > Options > Generating OpenType & TrueType > Kerning. They refer to the kern table (the old stuff) and the kern feature (the new stuff).

I hope that helps!

Regards,

T

I think we do get value from Kerning onscreen when we use ClearType, particularly when we render with sub-pixel positioning. Actually this was the original design intent for the ClearType fonts. I have some experience in this, as we included some device kerning in Palatino Linotype for onscreen black and white rendering. In the monochrome world the value of Kerning onscreen for normal text sizes was less obvious. You might get away with a To kern for instance, but lowercase kern pairs were problematic, as you have to move 1 whole pixel to kern the letter pair, which is often way too much. In ClearType we have greater accuracy when placing the glyphs onscreen, and so the adjustment of the Kern value, is less than a pixel, which although not optimal still adds value in my opinion. Have a look at the enclosed screen shot, the font is Constantia, designed by John Hudson, at 10, 12, 14 point. We can of course debate the details, and some will say yes but it’s not perfect, on the other hand I see more value having the kerning working, than not.

The current plan with regard to the CT fonts is to ship with OpenType Kerning tables only

mike

In both OT PS (.otf) and OT TT (.ttf) there can be two kinds of kerning: OpenType "kern feature" kerning (which may employ classes or and/or plain pairs) and old-style "kern table" kerning (which only uses plain pairs). However, the relation between the types of kerning differs between the PS/TT flavors.

Principally, new-style OpenType Layout-aware applications (Adobe CS, Windows Vista WPF, Mellel etc.) use the OpenType "kern feature" kerning only in both flavors. If a font is an OT TT font and the old-style "kern table" is present, then old-style OpenType Layout-non-aware applications (Word, Quark, Flash etc.) will show the kerning specified in the table but if the old-style "kern table" is not present in the font, there will be no kerning. If a font is OT PS and the old-style "kern table" is present, old-style applications will show the kerning in the table, but -- and here is the difference -- if the old-style "kern table" is not present, the Adobe font driver (ATM, Windows 2000/XP built-in, Mac OS X built-in) will automatically convert the Western subset of the OpenType kerning into old-style kerning on the fly, so you will always get kerning for Western characters.

In FontLab Studio 5, you can explicitly decide (in Options / Generating OpenType & TrueType / Kerning) whether the font (either OT PS or OT TT) should or should not contain the OpenType "kern feature" and/or the old-style "kern table". In FontLab 4.6, OT PS fonts will never contain the "kern table" while OT TT fonts will always contain it.

Adam

I missed this thread when it started, so will just add a few belated notes on the kerning in the ClearType collection because it was sort of interesting. As Si and Mike note, the fonts all contain OpenType Layout GPOS kern feature kerning, rather than a 'kern' table. At one stage, we considered including both, the latter with a subset of kern pairs for the benefit of applications that don't handle GPOS kerning, but Luc de Groot reported various bugs he'd found in software encountering fonts with both the GPOS and old kern table kerning, so we have not pursued this option.

The OTL tables for all fonts were written using VOLT, not FontLab, for a couple of reasons. Firstly, all the fonts contain lookup types not supported by the FDK/FontLab code. Secondly, Microsoft wanted VOLT files for source management.

When we began the project, I had already made some fonts using VOLT, and was concerned that it is far from an ideal tool for developing kerning data. There is no way to preview strings of more than two glyphs in the pair adjustment GPOS user interface. So Paul Linnerud from the Advanced Reading Technologies team and I designed a process to convert kern pair data from an AFM file into VOLT lookup syntax. This would enable the designers to do their kerning in FontLab (or FontStudio in Luc's case), and then convert to VOLT source data. At the same time as Paul and I were working on this, I mentioned to Sergey Malkin at MS Typography that we were looking for a kerning converter. To my surprise and joy, Sergey very quickly delivered a small tool, kern2volt, which was much simpler than the complex syntax conversion I was working on with Paul and produced a smaller, more effiecient lookup. The kern2volt tool examines a kern table and produces an efficient GPOS lookup with classes based on shared values. In the end, we used this tool for all the ClearType fonts. It is now available for download from the VOLT user community on MSN.

During this process, Luc and his associates in Berlin were busy working on their own kerning conversion method, which took a third approach involving writing VOLT glyph groups to match FontLab kern classes and then writing corresponding kern lookups. We used Luc's VOLT kern sources made in this way for the early test versions of his Calibri fonts, but discovered a bug in the TT rasteriser in the earlier versions of InDesign that caused the font not to render correctly. This was already fixed in the more recent versions of InDesign, but we decided to convert Luc's kerning data using the same kern2volt procedure as we had the other fonts in the collection.