[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: OFL license analysis

Mark Rafn <dagon@dagon.net> wrote:

>>> don@debian.org wrote:
>>>> Debian decides to distribute works containing your font. The
>>>> original upstream disappears. A bug is discovered in the font, and
>>>> Debian needs to fix it.
>> On Sun, 29 Jan 2006, Marco d'Itri wrote:
>>> Yes, and this is considered a feature. Usually existing documents
>>> should not break because a font is changed, even if this fixes a
>>> bug.
> On Sun, 29 Jan 2006, Don Armstrong wrote:
>> The same argument applies equally well to programs. We should be
>> intelligent enough in our fixing of bugs in fonts not to break
>> existing documents, 

That's plain impossible.  A bug in a font could be a wrong
kerning. Kerning means the hints in the fonts that indicate that in a
particular combination of letters, two adjacent letters need less (or
more) space between them than their size would dictate. For example 
the letters "VA" must be shifted a little towards each other, otherwise
it will look nearly like a whitespace between them.

If there is such a bug in a font, changing it implies that the
word-wrapping algorithm has a high chance of finding a different
solution, and thus change existing documents.  There's simply no
solution that allows both for stable docments and for bug fixing.

The lmodern fonts, for example, are still under development, and they
warn the user and reserve themselves the option to change things like
the kerning.  But if you've got a font that is in wide use and regarded
as stable, changing the kerning is a design decision and should in fact
change the name under which the font is available to the user and to

>>just like we should be intelligent enough in our
>> fixing of bugs in programs not to break existing scripts.
> This discussion seems to have gone into the weeds about WHY someone
> would want to make a change and whether Debian is able to make such
> changes reasonably.

Well, only in part.  A font that you can't rely on is mostly useless...

> Name-change requirements are
> acceptible (barely) on the package name, but not API identifiers, and
> that includes filenames that are part of an API.
> What ever happenened to the LaTex license, by the way?  That had the
> same non-freeness.

You seem to have a different opinion on this as the debian-legal people
who negotiated with the LaTeX project and together developped the LPPL
version 1.3.  In this version, the requirement to change file names (or
LaTeX package names) was dropped, probably because it was regarded as
"too non-free", whereas it was decided that any changed version should
indicate that it was changed in the version string that is (or to some
degree only will be) part of the API.

> It seems a clear test: if I can't distribute a changed version that
> can be dropped into a system without changing other software,
> it ain't free.

You can never distribute a bugfixed version of a font with the same name
(identifiers, ...) and, without changing other software, get the same
system behavior.  That's not a question of freeness, it's a technical

Regards, Frank
Frank Küster
Single Molecule Spectroscopy, Protein Folding @ Inst. f. Biochemie, Univ. Zürich
Debian Developer (teTeX)

Reply to: