Re: How should one handle condensed etc., variants?

[Rebecca and Rowland <rebecca@astrid.u-net.com>]

Melissa O'Neill writes:
[snip useful bits]

Thanks :-)

>Rowland continues (reformatted):
>> I'd say the NFSS needs at least three more selection axes:
>>
>>     weight: to allow for the selection of bold, light, etc., separate
>>             from the informal/other variants there might be.
>>
>>     angle:  to allow italic/slanted/upright/etc to be selected separate
>>             from shape (small caps, or whatever)
>>
>>     width:  to allow condensed/extended to be selected separate from series.
>
>Hmm, I'm not sure about having `angle', but I'd agree that if we were
>redesigning NFSS, I'd advocate separating `series' into `weight' and
>`width'.

I've had some direct email from Barbara Beeton (bnb@math.ams.org) which
touches on this - she reckons (and I'm inclined to think she's onto a good
idea here) that `italic' should be specified as a shape, with a separate
`slope' axis dealing with whether or not it's forward sloping, backward
sloping, or upright.  I've had some thoughts on this myself - I've tacked
the more relevant ones onto the end of this email.

> (I believe that we could then do away with series, since it
>only represents a combination of weight and width.) I'd also recommend
>adopting a *set* of attributes for shape, including italic, roman, serif,
>sans-serif, small-caps, swash, old-style digits, etc.

Hmm.  Any idea how you might do this?  One might, I think, retain the
`series' axis to cover things like old-style digits (for example) - you
could specify a variant using old-style digits, or any other arbitrary
thing like that, using series, while leaving the other design axes `clean',
and covering a clearly-defined range.

>But, for all this, I think it may be a little late to be trying to revise
>the font selection scheme.

For the moment, perhaps, but a new fount selection scheme to replace the
current NFSS is clearly needed.  I think it'd be worth discussing it here,
so a coherent proposal can be put to the LaTeX 3 people that deals with all
the bits that people think need dealing with.

After all, the NFSS itself predates LaTeX2e; why shouldn't the LaTeX 3
fount selection scheme be implemented as a bolt-on goodie for LaTeX2e?

>Rowland also writes:
>> [My proposal] doesn't take into account old style variants, which I've
>> never used and never really thought about.
>
>Well, from what I've learned about typography, they seem to be pretty
>essential if you're setting numbers in a text paragraph. The `titling
>figures' usually found in fonts are much to big to do anything other
>than accompany ALL CAPS titles or fill tables of digits.

This does seem to make sense.  Maybe I'll have learnt enough about fontinst
to create my own old-style digit founts soon - I've always thought they
look better in text, but I've never been able to use them conveniently.

btw, taking Thierry Bouche's <Thierry.Bouche@ujf-grenoble.fr> recent
message into account, how about these axes?

encoding/family/series/shape/slope/width/weight/case/ptsize

(I'm having grave misgivings about this, because I belong to the minimalist
school of thought, but can anyone see a way of reducing the number of axes
and still allowing proper flexibility?)

>Thoughts welcome,

<cackle>  If you've got any spare, my poor, empty brain could use them -
it's not seen a thought in *weeks*.

Anyway, about this italic/slope business:

Barabara Beeton (bbeeton <BNB@MATH.AMS.ORG>) wrote:

>now, as to the axes you propose:
>
>    So rather than:
>
>       encoding, family, series, shape (now)
>
>    I suggest:
>
>       encoding, family, series, weight, shape,  angle,     width
>
>    e.g.,
>
>       T1         ptm   informal m/bx    sc/n   it/n/sl/ui  cn/ex
>
>    Anyone care to comment?
>
>there's still some conflation here of shape and "angle" (i prefer
>"slope").

I think that *is* a nicer term; but I might be wrong ;-)

>  italic glyphs often have a different shape than the
>corresponding "normal text"; this is true of the latin a and g,
>also of the cyrillic v, t and others.  you can see this in computer
>modern -- cmsl is simply an inclined roman, and cmu is an upright
>italic.

Good point - I'd spotted this, but I hadn't come up with any ideas on how
to deal with it.

>  (the national geographic magazine uses an italic font for
>picture captions; when these include a proper name -- of a ship,
>say, or a biological species -- these are in the same *shape*, but
>upright.  so upright italic is not just a "donald knuth thing".
>i seem to remember being told that the font used in the geographic
>was created by matthew carter.)  thus i suggest that "italic" be
>applied as the shape, and upright, forward-sloped and backward-sloped,
>and perhaps even the degree, be used to designate the angle.  (if
>you've looked at the computer modern parameters, you might remember
>that the slope for some are 1/6 and for others, .25; unfortunately,
>i don't remember which is which at the moment, and i'm too rushed
>to check.)

Well, the exact details don't matter too much at the moment; it's the broad
sweep that counts.  One thing that does concern me, though, is how to
implement LaTeX's very nice switching between italic and upright on use of
the \emph{} command.

Taking on board your idea about the possibility of different angles of
slope; I'm not sure how this could be used easily if one is to be able to
select a forward-slope (say) with a single specifier.  One could use, say,
f, u, or b for forward, backward, or upright slope, but I'm not sure about
specifying the precise angle.

One way out might be to have two sequences of slope specifiers.

1) f(orward), b(backward), u(pright)

2) f45, f60, f30, b45, b60, b30 (etc.)

The idea being that the primary fount definition is given for the precise
angle and direction of lean; and a silent substitution is declared for F,
B, and U.  This way, you get to discriminate between different slopes (if
this becomes needful); and you also get to specify forward, backward, or
upright as usual.

Maybe not the most elegant solution, but one that might work.

As for the problem of \emph{}, I suspect that it'll be necessary to
introduce a new sort of thing in fount definitions to set up a mapping
between founts so LaTeX knows what to switch between.

I'd suggest (for simplicity's sake) a mapping from one instance of shape
and angle to another, done on the basis of `if you're using this
combination, switch to this one for emphasis, then back again' - not a
reversible specification

So, for example, you might specify something like:

\DeclareEmphPair{n,u}{it,f}
\DeclareEmphPair{it,f}{n,u}

Note that you'd have to do it like this, because of:

\DeclareEmphPair{n,f}{n,u}

(you can't have the emphpair reversible, because you'd normally want to
switch to italic shape from roman; but you need to switch to *something* if
you're slanted.)

And then there's:

\DeclareEmphPair{sc,u}{scit,f}

This shows something that I'm not entirely happy with: the way you've got
to specify a separate scit shape, for example.  It somehow feels wrong,
even though it might work perfectly well and logically.

But, taking Thierry Bouche's <Thierry.Bouche@ujf-grenoble.fr> suggestion
for a case axis into account, the scit mapping would be redundant.

But I suspect it'd still be necessary to provide some kind of mapping -
does anyone see a way round this?  I don't like my suggestion, but it seems
necessary to me.

Well, that's my bit for now.  If anyone's survived reading that, do you
have any thoughts?

Rowland.