Ummm... probably pretty useful.
but not necessary.

that well known competitor ignored a lot of unwritten "type ethics" by modifying and selling your outlines, even if he was technically free to do so.

To avoid this problem, I'd restrain from publishing the source files along with the font, but make them available upon request. I'd guess that the average person who downloads a free typeface doesn't have much interest in or use for the source files. Publishing the source files is gonna mean different things to different people.

for the average person it might mean
"I'm a well developed typeface, download and use me!"

for the type / graphic design hobbyist it might mean
"Hey! Play around with me and make something new!"

for the (shady) type / graphic design professional it might mean
"Hey! Play around with me and make something new and sell it!"

and for the type design benefactor it might mean
"I'm a well developed typeface, develop me more!"

So I guess what I'm trying to get at here is, yes, it would help to have those source files for "Bob". But if you release "Alice" under the pretense of "Do whatever the hell you want!" then you indirectly brought the whole "Bob" situation upon yourself. But realistically, if they modified "Alice" enough, then the question is irrelevant to the previous states of "Bob" and you're really asking "How useful would it be the have the source files when asked to expand ANY typeface family." To which the answer will most likely always be "Probably pretty useful."

Perhaps a tiny real-world example?

I modified Charis SIL and used the resulting font to set several books for university presses. Modifications were: old-style numbers, character spacing (sidebearing/kerning modifications), condense the italic a trifle, and replace the mark and mkmk feature with a ccmp feature (I use FontLab only).

"Our" Charis is not for sale. I believe the SIL license forbids this, but in any case, I wouldn't do it. I have given it to one press to use in a series they are publishing.

Well, it wasn't Bob -- Charis was pretty good to start with; itself a modification of Mathew Carter's Charter. Another SIL font, Gentium (revised), is now in beta testing. They make the source material available. It is of no use to me; I'd write my own features anyway. And Gentium (not Gentium Book) is a touch too light for bookwork with high dpi, ctp offset printing.

This doesn't directly answer your question. As I understand it, your model was that a *reasonable* font was turned into something of beauty by modification, but not all characters were offered. What would it take to get the missing characters? And of course, it would take the same skill that turned the font into a thing of beauty. How would the source files, beyond what FontLab will extract from the OTF, help with that?

It depends on how the font were done, if the glyphs were mostly drawn by hand, OpenType stuff done manually etc., then the working sources are of limited use. But if the weights were interpolated from masters, OpenType was (semi-)automated with scripts, then the sources are pretty important, but in either cases it is not necessary, you can still extend the font but not as efficient as if you have the sources.

Andrevv,

that well known competitor ignored a lot of unwritten "type ethics" by modifying and selling your outlines, even if he was technically free to do so.

The whole point of publishing a libre font is to encourage people to modify and sell font outlines :-)

if you release "Alice" under the pretense of "Do whatever the hell you want!" then you indirectly brought the whole "Bob" situation upon yourself.

No libre license says "Do whatever the hell you want!" - libre fonts come with a license that is as firm as the EULAs normally used for fonts, and it allows certain things, requires certain things and prohibits certain things.

Really the main libre license is the SIL OFL v1.1 - and this is the license preferred by all the big libre font projects.

- - - - -

Charles, thanks a really great example! Thanks for telling the story :-)

"Our" Charis is not for sale. I believe the SIL license forbids this, but in any case, I wouldn't do it. I have given it to one press to use in a series they are publishing.

Making something 'for sale' involves 2 separate things: offering to distribute the work to the public, and charging money for distributing it.

The SIL OFL allows charging money for distributing OFL fonts when a program is included alongside the font; anyone can include a "hello world" program to sell the fonts. Indeed, anyone selling computers (like One Laptop Per Child, or Dell) with Ubuntu or any other libre operating system installed is selling the fonts.

No libre license requires publication though - the private right to modification, which you are exercising - is one of the major motivations for libre licensing! :-)

As I understand it, your model was that a *reasonable* font was turned into something of beauty by modification, but not all characters were offered.

There is a difference between

Not quite, the idea is that the first font is just as beautiful as the second, but is a particular style, and covers a lot of weights and characters. The second font is a different style, and covers less weights and characters, but retains many of the characteristics of the first one. So someone would like to fill out the second one to match the number of weights and character coverage of the first - but don't have the VFBs.

How would the source files, beyond what FontLab will extract from the OTF, help with that?

The interpolation masters can be quite different from the distributed installable files; they have PostScript curves, and the distributed files are TTFs with Quadratic outlines - so no on can take the regular and bold and interpolate between them, without converting outlines back to PostScript and going through and cleaning them all up to prepare them for interpolation again....

This seems really annoying and a total waste of time to me, but, its not totally impossible, anyone competent could do it in a day or so - right?

So.... what I'm really wondering, Is if there is anything in a set of made-for-installation OTF or TTF files that would take weeks to recreate for convenient modification?

- - -

you can still extend the font but not as efficient as if you have the sources.

Khaled, yes... I'm wondering, what is the upper bound on how bad the inefficiency can get? :-)

Would it be good to require libre font distributors to provide source on request?

- - -

To be clear, my posts here are my personal opinion only, and do not represent the views of any of my consulting clients.

Hi, my own type foundry VTF (www.vtf.fadebiaye.com) has a specific policy (http://www.fadebiaye.com/type/vtf_policy.html) about that specific matter.

We do not develop font, we provide fonts under CC BY SA licence or under GNU/GPL. We do provide sources and we allow people to modify and resell it, but they have to do it under the same licence, unless they ask me directy ; in that case, as the author of the font, i can make the font private.

Of course, people might not respect that deal ; but in any case, they would have stolen an undeveloped typeface. But who cares?

I must add that i do experimental directly on computer ; with a specific grid-aided blueprinting in first place ; so of course, source files give great aid.

@Dave:
Khaled, yes... I'm wondering, what is the upper bound on how bad the inefficiency can get? :-)

It can be very limiting, for example, in the font I'm about to release, I strip glyph names (not required for TTF fonts) and lookup names don't get exported either, so if someone is to alter some of the lookups in the font without my working sources it'll be a very laborious effort. Further more, I internally use kerning classes that are "exported" at build time to kerning pairs, editing thousands of kerning pairs is no fun. Even worse, for fonts built from Matafont like sources (many of GUST fonts), good luck editing 40 font files or so per family that were actually generated from a single Metafont master.

So in general, I'm more with providing sources and build scripts (if any), it won't harm.

Dave, I think you hit upon an interesting phenomenon. Many publishers of Open Source fonts intentionally (or unintentionally) fail to be "open" with actual "source". Perhaps in the case of VFB files these can be reverse engineered from the binary? But in the case of VTT hinting private tables, and to a lesser degree VOLT OpenType tables it's harder said than done.

Sii, Adam Twardoch has talked publicly at conferences this year (I forget which one) about making FontLab files interoperable for libre font tools. I speculate that could be publishing the VFB format specification, or perhaps making a documented variant of UFO ("XFO"?) the default file format for FontLab Studio 6 and providing way to upgrade VFBs to XFOs in FL6.

Khaled, I think when more tools like "Proteus" are available, the need for sources will become even more aparent. The next generation of font tools all seem to take on the parametric ideas of METAFONT, and the type design tool R&D that FontLab demo'd at ATypI this year was VERY impressive, being interactive and DESIRABLE. Your points about the current tool's ability to make the installable fonts very different and annoying if interesting. I wonder if a tool could be made to deduce and reconvert 'flat' kerning pairs into a hierarchical kerning class system... and the same for OpenType lookups, that seems much much harder though.

Frank, the GNU GPL does require sources to be made available to users on request, which is why its my favourite font license. However, CC-BY-SA does not, and I think its better to use the SIL OFL v1.1 instead of CC-BY-SA for fonts. Fonts are not 'content', they are much more like programs (and contain programs.)

Thanks guys!

- - -

To be clear, my posts here are my personal opinion only, and do not represent the views of any of my consulting clients.

The lack of a requirement to release all modified source materials for derivative works is in my opinion a major drawback of the Open Font License (OFL). The OFL lacks a major part of the open source culture, which is that someone should not need to reverse engineer or start from scratch when modifying an existing project (program, font, etc). For future releases of my original fonts, I am looking for an alternative to the OFL.

I don’t blame font designers for not releasing source files. If you release the VFB files open-source zealots complain that you’re not using FontForge and insist that the project be switched to UFO, and once it is, they contribute nothing. If you release the UFO files you get emails from people who can’t RTFM about UFO. I still get clueless emails about OpenBaskerville even though it’s not even an active project.

I don’t blame font designers for not releasing source files. If you release the VFB files open-source zealots complain that you’re not using FontForge

May be I'm "open-source" zealot, but releasing plain TTF files is much more useful than VFB files, since the later is a proprietary, undocumented file format that only applications written by one company can read an edit.

If an open source project is not gaining outside contributions, then I take that as their failure to attract contributors, opening up your project won't magically make people willing to contribute to it. Take DeJaVu font project as an example of successful open-source font project.

>Take DeJaVu font project as an example of successful open-source font project.

So as a type designer you have to develop a font for a browser maker and then remove the manual hinting, change the name and license it to a desktop shell producer, and have them open source it? That seems a bit much for most type designers wanting to get into the Open Source font game? ;-)

Dave (abattis) -

This all sounds very much like a discussion you and I and others have had many times before, so I'm surprised to hear you bringing it up yet again. But for the benefit of others here...

It is surprisingly difficult to define what files constitute complete 'source' for fonts. is it every file used in development, such as scanned source of drawings before optimisation? Is it every file necessary for 'building' the font, including every small Python script used to, say, work around a FontLab bug? Does it include app-specific prefs files, such as FontLab .nam and .enc files? Do they need to be in an open, non-proprietary format, so that anyone can use them? Does it do anyone any good if the files are there but without documentation on how to use them?

Fonts are not like other kinds of software that can easily be compiled from a 'source' in a documented, automatic system that needs only a one-button push. We and many other foundries are working hard to get closer to this, but only a few really succeed, and I suspect that almost none of those in-house build systems could easily be reproduced by others. The build system for our Roman fonts (Gentium, Charis, etc.) is horribly complex. We've worked very hard over the past three years to make it simpler, but there is a natural complexity in such massive projects. If we were required to bundle 'source' files with these fonts (according to one strict definition of 'source' - a complete reproducible build tree) there is no way we could have done it.

For most purposes, the resulting font itself can be considered sufficient 'source'. The outlines, data tables and even smart font behaviors are there and can be modified and improved. Hinting and complex smart font routines can be more opaque, so there's clearly usefulness in a richer set of source files. It can be difficult, however, to produce that richer set, and requiring us to do so would be prohibitive.

Most importantly, if you require foundries to produce 'source' upon request, you need to produce an objective definition of what that is. IMHO, there is no such definition that would be workable. We tried very hard to come up with something like this when developing the OFL, but kept finding big problems with anything we tried. In the end, the font file itself became the basic minimum standard of 'source', and we any many others felt that was sufficient to call the OFL a free/libre license.

We do feel it's important that foundries share as much as they can with others. For example, although there's no way we can currently reproduce our build system on a user's computer, we do provide additional 'source' files that would be reasonably useful for people wishing to build derivatives, such as FontLab databases, OT-CFF files (PS curves in non-proprietary format), anchor definitions, Graphite code, etc. We even produce some of these (such as OT-CFF) even though they are not needed for our build systems.

In your specific case, the solution lies not in codifying a source definition in law, but in developing a relationship with the guys who produced "Bob". Ask if they can share their additional work with you in a useful way. If they don't respond well, see if a mutual friend (Ted?) can talk to them on your behalf. Offer to do some extra work to help make their work more broadly useful. Keep in mind that most derivatives are produced by someone in their spare time to meet a specific purpose. They may not have the technical ability or time to produce a richer set of development files.

To require complete 'source' is a dead end in many ways. It is impossible to define, might require expertise and resources many font designers don't have, and would only stifle the development of free/libre fonts.

BTW, Dave is referring to the Proteus system under development at Fontlab Ltd. with partnership with Yuri Gordon. It's a parametric type design system. The talk announcement can be seen at:

http://www.atypi.org/03_Dublin/40_timetables/preface/view_presentation_h...

and I believe that we will publish a presentation at some point soon for people to look at (those who haven't been at ATypI).

Also, the talk has been recorded by ATypI, and I think the video recording will also be published, though I don't know the specific details.

Best,
Adam

So as a type designer you have to develop a font for a browser maker and then remove the manual hinting, change the name and license it to a desktop shell producer, and have them open source it? That seems a bit much for most type designers wanting to get into the Open Source font game? ;-)

Non of this is related to DejaVu project, I'm referring to the project and its contributer friendliness, which has nothing to do with how the fonts were open sourced.

>I'm referring to the project and its contributer friendliness, which has nothing to do with how the fonts were open sourced.

I'd argue the opposite. The work that Jim Lyles and co. did directly contributed to the contributor friendliness of the project.

Sorry for going off-topic.