Re: urg.

[Han-Wen Nienhuys]

address@hidden writes:
> >
> >  1. Type1 fonts have to embedded binarily.
> >
> >  2. Type1 fonts have to be interpreted to extract glyph widths (with 
> > freetype?)
> >
> >  3. The Type1 has to be read, to determine byte offsets of
> >    different sections
> >
> >  4. There is no glyphshow operator in PDF (damn). I believe that
> > Unicode is well supported for text, but for emmentaler, we have 
> > to something with CMaps, CIDs and other stuff I don't understand.
> >
> > I wonder whether this is a smart choice to try out; direct PDF seems
> > like more work than I anticipated.
> 
> That sounds as a set-back, but I do not have a clear picture of the
> size of the problems.  If we could estimate the amount of work that
> each problem should take, it would be easier to choose between PDF
> output and fontconfig support for gs.  Choosing the gs route could
> mean a problematic support on most platforms, but we could try to
> investigate that too (ask Red Hat, Debian, Cygwin, eg.).
> 
> It seems that some (1. binary embedding, 4. contstruct
> glyph-name->charcode mapping) are technical questions rather than
> glyph-name->things that require much work.


>  As for 2. exact widths,
> we already do that for whole strings (only need scm wrapper?); I do
> not understand what 3.  means.

The trouble is that embedding fonts in PDF needs to follow certain
standards: some info has to be filled in, and to do that, we have to
read the TTF/Type1 font and take it apart. I think that's not our
expertise, so we shouldn't be doing it.

I'm certain we can produce the PDF backend. It's just that I thought
it would be a convenient way to deal with TTF fonts. This might be the
case, but then we have to deal specially with Type1 fonts once again,
which makes it an unattractive solution to the TTF problem.

Oh, of course I forgot another option: we can construct a Type42 font
ourselves. Reading the Type42 spec the  complications being

* that we have to take apart the TTF slightly, because the sfnts table
has to be inserted in correctly-sized chunks 

* we have to extract the glyphname -> index table

* we have to generate an encoding vector (I expect that we can just
  put adobe standard encoding here).


Actually, reading the spec, it doesn't seem all that much work. Taking
apart the TTF might already be done by FreeType, we just have to write
the tables in the correct order to a file. Werner? 

-- 

 Han-Wen Nienhuys   |   address@hidden   |   http://www.xs4all.nl/~hanwen