Re: [XeTeX] Should xelatex have its own kernel? (was: "Conflict between xunicode and fontspec?")
Ross Moore
On 08/02/2008, at 5:24 AM, Joel C. Salomon wrote:
> On Feb 7, 2008 3:28 AM, François Charette <firm...@ankabut.net>
> wrote:
>> U+2007 -> ? "figure space"
>
> A space the size of a numeral (if using lining figures; if the font
> only has proportional figures I'd guess '8').
>
> Although... mightn't the font include the width of these various
> spaces? I'd suggest using the values from the font, if such exist.
That is a good point.
For many of those special spaces (e.g. the math-spacing ones)
TeX will have rules that handle those situations.
However, this begs the question...
... what to do if they appear within document source?
For example, if a piece of LaTeX source has been constructed
by copy/paste from an existing PDF or other file, then these
characters may well be included.
Should XeTeX treat them as a normal space tokens, then let
the typesetting context determine what to do ?
Or should the meaning of these space tokens be adapted to
the current font, using TeX's \fontdimen parameters
for the font(s) specified for typesetting mathematics?
Indeed, should XeTeX be smart enough to *insert* these
spacing characters into the output that it creates?
This is particularly relevant to the typesetting of
mathematics, and it would indeed be a departure from
the way TeX currently works.
Against this is the question of how do "smart" math fonts
handle the spacing? (e.g. the STIX fonts).
Aren't they supposed to know the correct amount of space
to put around operator and relation symbols?
So adding these characters explicitly would be either
redundant or just plain wrong.
More questions.
Interesting thread.
>
> --Joel
Cheers,
Ross
------------------------------------------------------------------------
Ross Moore ro...@maths.mq.edu.au
Mathematics Department office: E7A-419
Macquarie University tel: +61 +2 9850 8955
Sydney, Australia 2109 fax: +61 +2 9850 8114
------------------------------------------------------------------------
Will Robertson
> On 08/02/2008, at 5:24 AM, Joel C. Salomon wrote:
>
>> On Feb 7, 2008 3:28 AM, François Charette <firm...@ankabut.net>
>> wrote:
>>> U+2007 -> ? "figure space"
>>
>> A space the size of a numeral (if using lining figures; if the font
>> only has proportional figures I'd guess '8').
>>
>> Although... mightn't the font include the width of these various
>> spaces? I'd suggest using the values from the font, if such exist.
Without knowing the reasoning behind this glyph, I'll go out on a limb
and predict that this space character is designed to be used in
tabular material with fixed-width numbers, when a usual space wouldn't
necessarily be the right width for the text to align vertically
between lines.
> For example, if a piece of LaTeX source has been constructed
> by copy/paste from an existing PDF or other file, then these
> characters may well be included.
>
> Should XeTeX treat them as a normal space tokens, then let
> the typesetting context determine what to do ?
I think this is an open question. Many of the space tokens in unicode
have unambiguous meanings so if you see a thinspace in the source then
it makes sense to keep it for the output. Especially for text
processing.
> Indeed, should XeTeX be smart enough to *insert* these
> spacing characters into the output that it creates?
> This is particularly relevant to the typesetting of
> mathematics, and it would indeed be a departure from
> the way TeX currently works.
I think we can safely assume that the TeX community knows more about
mathematical typesetting than any font designer (with a couple
exceptions). Therefore I'd say it's reasonable not to use the unicode
characters for these spaces when PDF is the output mode. Most fonts
don't even have most of the characters, right?
> Against this is the question of how do "smart" math fonts
> handle the spacing? (e.g. the STIX fonts).
> Aren't they supposed to know the correct amount of space
> to put around operator and relation symbols?
> So adding these characters explicitly would be either
> redundant or just plain wrong.
I don't believe there is any font that smart yet.
Although I don't think it's impossible... But what font designer would
bother, when you need a typesetting program to do at least some of the
work for you?
I mean, you could approximate good maths typesetting with plain
unicode text with all these spaces and so on, but you'd still not be
able to do superscripts or subscripts, or limits, or integrals that
change size, or square roots, or ... so what's the point?
Will
Joel C. Salomon
<Re. U+2007 Figure Space>
>
> Without knowing the reasoning behind this glyph, I'll go out on a limb
> and predict that this space character is designed to be used in
> tabular material with fixed-width numbers, when a usual space wouldn't
> necessarily be the right width for the text to align vertically
> between lines.
Correct; from <http://fileformat.info/info/unicode/char/2007>:
> space equal to tabular width of a font
> this is equivalent to the digit width of fonts with fixed-width digits
> > Indeed, should XeTeX be smart enough to *insert* these
> > spacing characters into the output that it creates?
> > This is particularly relevant to the typesetting of
> > mathematics, and it would indeed be a departure from
> > the way TeX currently works.
>
> I think we can safely assume that the TeX community knows more about
> mathematical typesetting than any font designer (with a couple
> exceptions). Therefore I'd say it's reasonable not to use the unicode
> characters for these spaces when PDF is the output mode. Most fonts
> don't even have most of the characters, right?
I've been favorably impressed by Murray Sargent's "Nearly Plain-Text
Encoding of Mathematics" (http://unicode.org/notes/tn28/); perhaps the
copy-and-pasteable text in the PDF file should be (mostly) compatible
with that.
Oh, and the space characters are *not* present in the utn28 encoding.
They are used only to select the spacing around operators &c.
(I could be wrong about all this, of course, but that's my reading of
Murray's paper. Will, you want to ask him directly?)
> > Against this is the question of how do "smart" math fonts
> > handle the spacing? (e.g. the STIX fonts).
>
> I don't believe there is any font that smart yet.
> Although I don't think it's impossible... But what font designer would
> bother, when you need a typesetting program to do at least some of the
> work for you?
*Some* of the spacing options are set in a good maths font.
(Generalizing from -- what, three? instances.)
> I mean, you could approximate good maths typesetting with plain
> unicode text with all these spaces and so on, but you'd still not be
> able to do superscripts or subscripts, or limits, or integrals that
> change size, or square roots, or ... so what's the point?
Doesn't Cambria Math include different glyphs for subscript and
scriptscript sizes, and for constructing integrals and radicals of
different heights? But for copy-and-paste, they should be encoded as
'∫' or '√'.
--Joel
Will Robertson
> I've been favorably impressed by Murray Sargent's "Nearly Plain-Text
> Encoding of Mathematics" (http://unicode.org/notes/tn28/); perhaps the
> copy-and-pasteable text in the PDF file should be (mostly) compatible
> with that.
I do agree that the least amount of markup without losing information
is the best way to input maths. His encoding is a good system but
completely incompatible, at this stage, of how (La)TeX does maths.
> (I could be wrong about all this, of course, but that's my reading of
> Murray's paper. Will, you want to ask him directly?)
I'm unfortunately not in contact with him.
>> I mean, you could approximate good maths typesetting with plain
>> unicode text with all these spaces and so on, but you'd still not be
>> able to do superscripts or subscripts, or limits, or integrals that
>> change size, or square roots, or ... so what's the point?
>
> Doesn't Cambria Math include different glyphs for subscript and
> scriptscript sizes, and for constructing integrals and radicals of
> different heights? But for copy-and-paste, they should be encoded as
> '∫' or '√'.
Yes, it contains all that information. All I'm saying is that
information can't be expressed in a stream of plain unicode text, so
why should the font designer spend all that time inserting what is
essentially kerning information between a huge number of characters
that will only be replicated by the real maths typesetting program.
Not this is not the same as including in the font the sizes of the
skips that should be used! XeTeX and Word do use the metadata in the
font to apply the correct spacing rules. But this is different than
explicit kerning information being present between the individual
maths glyphs.
Cheers,
Will
Will Robertson
> Accordingly, I just got Office 2008 for the Mac, hoping that these
> capabilities would be implemented in it as well. Alas they don't seem
> to be there, and the Mac version seems to still entirely rely on
> Equation Editor. Did I miss something?
WHAT?!?! No, you didn't miss anything:
<http://groups.google.com/group/microsoft.public.mac.office.word/msg/17de2d0cf51a8208
>
I can't believe this. I thought they pretended to have broad feature
equivalence between Mac and Windows. The compatibility nightmare that
this could cause some people...
Well, obviously there wasn't enough time for them to implement this
new feature. I mean, they did have the Intel compatibility thing to
worry about. And it's not like maths typesetting is a feature that the
majority of people will miss. But still.
Will