Re: Margin kerning

[Ludovic Courtès]

Hi,

address@hidden (Jeff Kingston) writes:

> Ack.  I'll look over it more carefully when I come to set
> up the next release.  You certainly have made it easy for me.
> My only question is about your definition of a margin-kernable
> character, as one that is less than 3/4 the x height.  Would
> this work for continental quotation marks, which from memory
> look like this:
>
>     <<Margin-kern me!>> said the chevrons.

Well, it depends on the font being used.  Consider the top-left of page
4 of the two following documents:

  http://www.laas.fr/~lcourtes/software/lout/ubimob2005-paper.pdf
  http://www.laas.fr/~lcourtes/software/lout/ubimob2005-paper-pa.pdf

The former uses the Latin Modern Roman font while the latter uses
Palatino.  With Palatino, the French "guillemet" `<<' gets
margin-kerned, while it doesn't get kerned in LMR, because LMR's glyph
for guillemets is too high.

Now, the question is: why 3/4th instead of 1 or one half of the height
of `x'?  Intuitively, I initially considered that `x' is usually the
smallest glyph, and that any glyph strictly smaller than `x' should be
eligible for margin-kerning.  But it turns out that, in Palatino, the
glyph for `v' is slightly smaller than that for `x'.  As a consequence,
`v's would get margin-kerned, which looks odd.  3/4th of the height of
`x' looked like a reasonable trade-off.  :-)

> Also you've limited it to the last/first character of a word.
> What if a word contains several last/first characters that
> satisfy the definition of margin-kernable?

Good point.  Indeed, the example above (in the Palatino case) and the
small test example both exhibit such a situation.  However, I don't know
what the best thing to do is.  In "Micro-typographic Extensions to the
TeX Typesetting System" [1], the author doesn't seem to mention this.
The examples therein don't exhibit this either.  But it seems that
"character protruding" is shown as a technique that applies to only one
character.

Intuitively, I doubt applying margin-kerning to more than one character
would be a good idea, because it may lead to having characters too much
protruded.  Likewise, it is debatable whether ellipsis should be
margin-kerned (as in the small example I posted).  In [1], the author
advocates that each glyph should have its own "protrusion factor"
determining how much it should be shifted outside the margin.  But then,
this is getting more complicated, especially since AFM don't provide
such information...

Thanks,
Ludovic.

[1] http://www.pragma-ade.nl/pdftex/thesis.pdf