[Pkg-fonts-devel] Common script for conversion of font files

[Rogério Brito]

Hi, Daniel.

On Feb 15 2011, Daniel Kahn Gillmor wrote:
> On 02/15/2011 03:39 PM, Rogério Brito wrote:
> > ... a BIG but is that we should *not* be recompiling fonts from their
> > sources, since we can introduce many, many problems (floating pointing
> > errors are just the beginning) and ruining the layout of the fonts.
> 
> I'm sorry, but i disagree pretty strongly with you on this point,
> Rogerio.  Debian is about free software, which includes the ability to
> modify the software.

Yes, I do "get the point" of what Free Software is (BTW, I am not one of
those that call Free Software "open source", so that you may see where I
come from).

I am *not* against having "buildability" of the fonts or whatever. I am
against not preserving the fonts.

> Making sure that we have the toolchain available (and working!) to
> successfully recompile fonts from source (their preferred form for
> modification) is as important as making sure we can successfully
> recompile other software from source.

I know all about that. The problem here is one that many people have been
fighting against, namely, numerical stability, given the limited memory that
we have in our current computers.

> These sound like bugs to be filed against the relevant parts of the
> toolchain, not reasons to avoid the toolchain altogether.

Unfortunately, it is not *possible* to have better toolchain here. I am not
talking here about feasibility or whatever. I am talking here about *real*,
mathematical possibility.

I am sure that you have sufficient mathematical background to understand the
very fine issues here, but I can provide you with some proofs or references
to what I am saying here.

> Towards a better free font infrastructure,
> 
> 	--dkg

We are talking here with a problem of approximately the same nature of GIMP
having only 8 bit/channel and the criticism of people that 8 bits are not
enough.

But here, perhaps, things are magnified, since we can't even represent some
"reasonable"

In other words, we won't have the infrastructure needed for that, not today,
not in a million years, not whenever. Only if we change our computational
models.

Summarizing, I am not against having the fonts built from source: just do
that in the package building phase and, say, use the command line utility of
fontforge to compare the result with what upstream gives us. If they are the
same, throw them away and ship upstream's version.

A recent case:

http://bugs.debian.org/609094#10
http://packages.qa.debian.org/f/fontforge/news/20110118T151726Z.html

Some recent commits from fontforge, just in case (please, notice the
freshness of the patches):

https://github.com/rbrito/fontforge/commit/2da94b7f69ed28cd234302f8d12626454b6fb6cb
https://github.com/rbrito/fontforge/commit/3233b5feb3ee8ffbabdbbe6008aab9d094704856
https://github.com/rbrito/fontforge/commit/33269cd22e41819efa72259d1831855b2b8c6a00
https://github.com/rbrito/fontforge/commit/ff25eaab48a9000901d61d4aad013b773f5a5aa0
https://github.com/rbrito/fontforge/commit/021288177a21d0cf22f14d9bf587fc56fcfb63aa
https://github.com/rbrito/fontforge/commit/94666672ee33068f874a774144f2a7e820b41e07
https://github.com/rbrito/fontforge/commit/754dd4945d675e7f1ff2c5dc7468acb8b0150fa5
https://github.com/rbrito/fontforge/commit/74a08529be176d04be6e81a267fc453965765edf
https://github.com/rbrito/fontforge/commit/49d8b8159f8255a04581712f500680ae14c389fe

And we are not even talking here about kerning and grid-fitting...


Regards,

-- 
Rogério Brito : rbrito@{ime.usp.br,gmail.com} : GPG key 4096R/BCFCAAAA
http://rb.doesntexist.org : Packages for LaTeX : algorithms.berlios.de
DebianQA: http://qa.debian.org/developer.php?login=rbrito%40ime.usp.br