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

Re: [RFC] Stripping Latin range in fonts used by g-i installer?

On 08/01/2008, Christian Perrier <bubulle@debian.org> wrote:
> Quoting Otavio Salvador (otavio@debian.org):
> > > The udeb size goes from 220768 bytes to 183090 bytes
> >
> > Impressive!
> Im waiting for some ACK by others to propose the strip patch to the
> font maintainer.
> I propose ttf-sil-abyssinica as a start, with some attempt to make the
> "strip" patch less dependent from the package than it is in
> ttf-freefont (many things are hardcoded there). I may need some help
> by people more skilled in well-written shell scripting than /me

I am throwing again my 2 cents that I've been keeping in my pocket
since the G-I Extremadura worksession, the same cents that made me
waste huge amounts of time then while trying to explain it and
implement it:

Why not do an automatic strip of the font and only *keep* the
codepoints that are used, based on the existing translations? I know
this could be too drastic and we run into the risk of loosing glyphs
if we are not doing this properly, but with Davide's graphic
comparison scripts we could detect such situations.

In other words, the idea would be to do something like:
 * for each font we use (and we consider we should strip it)
    * out of all the languages using it, out of all the strings in all the udebs
       * make a compilation of all the used codepoints
       * make a mask with the unicode areas that should always be kept
for the current font (e.g. for ttf-dejavu: latin, latin-extended-A,
latin-extended-B, cyrillic)
          * remove everything that is not part of any of the
compilation or the mask

   * use the resulted font

* PROFIT! :-)

Note: I am not able to work myself on this, but I can assist with
shell script expertise and suggestions anyone willing to do it.

"Imagination is more important than knowledge" A.Einstein

Reply to: