Font adjustments (was Re: what's an em)

Gosh, CC-ing to www-font caused much more of a stir than
just sending to www-style!  Good!

Since there were so many messages on this since yesterday,
I'm not going to reply to each individually.

Let me point out a few things, to clarify the goal here:

1. I'm not aiming for nanometer precision here, just that the
   "perceived" size is "sufficiently" well preserved when font
   substitution is done, or when doing explicit font changes
   (often between different typeface families).  Since automatic
   font substitution is done out of control of of the web page
   author, we should try to keep the "surprise factor" as low
   as possible.  For respect of the web page author and reader.
   There are quite enough problems with this in practice to
   justify some kind of reasonable adjustment.  It is unfair
   to both web page authors and web page readers to tout
   "point size" as the only true and just way of giving the
   display size of textual matter.

2. "Funny", excessively swashy, or other 'non-ordinary' typefaces
   don't count as counter-examples.  I'm talking about ordinary
   fonts that e.g. web newspapers would try to use regularly. 
   Font substitution still takes place, and will be different
   depending on which fonts are installed on a particular
   web page viewing device.

3. The suggestion to use Åp height, and cap height comes from
   "real typographers".  It's not my invention (in case that
   helps make anyone think again).  And the original complaint
   about the uselessness in practice of the "point size" also
   comes from "real typographers".  (I gave some references
   yesterday.)  This was slightly before the "web era" (1980-ies,
   early 1990-ties), but the issue has become acute now with
   the web and the font substitutions that take place.  

4. You need to know the cap height anyway, when mixing e.g.
   Latin with certain other scripts. E.g. Devanagari which is
   aligned at a high baseline, about where the (Latin) cap height
   is above the low (Latin) baseline, or when mixing with CJK
   which is aligned at a central baseline, halfway between the
   high and low baseline. (There is also a math baseline one
   needs to keep track of!)

5. I have no intention whatsoever to deny typeface designers the
   use of the "square", as a design help.  It's just that that
   square is of no particular interest to the font USER (web page
   authors as well as readers), no matter how definitive it is.

6. I'd be happy not to look at actual bounding boxes for glyphs,
   if NOMINAL x-height, cap height, and Åp height are declared
   by the typeface designer and stored (in one way or another)
   in the font file itself.  (E.g. the cap height could be taken
   as the distance between the low baseline and high basline.)
   For these nominal sizes (or baseline positions) the typeface
   designer may take percievedness issues into account.
   By all means, store a nominal M width, as decided by the 
   typeface designer, explicitly in the font file too.

7. These days Latin text is mixed with just about any other modern
   script.  I'm sure that some kind of reasonable relationship
   with the notions mentioned above has been developed, or can be
   developed.

I'm sure to have forgotten to clarify something, but before you
yell, please read my suggestion for CSS 3 again
(http://lists.w3.org/Archives/Public/www-font/2000JanMar/0029.html).
I still think it's a reasonable suggestion.

  Kind regards
  /kent k