W3C home > Mailing lists > Public > www-style@w3.org > February 2000

RE: What's an em

From: Karlsson Kent - keka <keka@im.se>
Date: Tue, 1 Feb 2000 21:47:08 +0100
Message-ID: <C110A2268F8DD111AA1A00805F85E58DA6857C@ntgbg1>
To: www-style@w3.org, "'www-font@w3.org'" <www-font@w3.org>

Some references with quotes and comments; the translations
are mine, some of the quotes are not in English originally:

Goossens, Mittelbach, and Samarin. The LaTeX Companion, 1994.
ISBN 0-204-54199-8. Page 449, Table A.1, (La)TeX's units of length:
"em      Width of "M" in current font"

(It is actually closer to the width of a "W", it's the
'nominal' width of M...  Without changing the "size" of
the letters, the em in TeX grows wider for bold style. 
This does not change the line height, of course.)


Bo Berndal. Typiskt typografiskt.  [199x]. ISBN 91-7054-670-3.
Page 66, top, in the illustration (my translations):
"Metric 'grad' [size] = usually H-height in mm or tenth of mm."
"The p-height is normally the smallest recommended
[base]line distance for Swedish text."
"The old concept of "square" corresponds in principle
to the p-height measurement in width and height."
"p-height corresponds approximately to the lead type's
basic 'kgel' [height]."
Page 64, commenting text to the top illustration:
"Theoretical type sizes. To the left set in 1989 with 10/11 
Perpetua. To the right set in 1990 with 10/11 Perpetua.
The above nominally equal type sizes differ by 24%. One sees
that the fictuous point system does not work in practice.
The only correct thing to do is to have a measure that
can be checked with the printout, e.g. the cap height in mm."


George Flessa. Trycksaksboken. [199x]. ISBN 91-23-01348-6.
Page 44, text to the bottom illustration: "A square of
36 points (to the left) and a few letters of that size.
Compare the glyph images's real size to that of the square
(which corresponds to the size of the lead type)."
[Note that the 'square' is not an 'em square', which would
be smaller, which would not work well in lead...]


ke Hallberg. Typografin och lsprocessen. [199x].
ISBN 91-7136-432-3.  Page 74: "The square corresponds
approximately to the p-height squared." 
Same page further down: illustrates the EM with a
capital M (not the p height).


I know, the notion of an "em" is thoroughly confused (I will
not repeat the result of a web search I did some time ago, it's in
http://lists.w3.org/Archives/Public/www-style/2000Jan/0123.html.)


Looking through those clippings, and a few more typography
books often gives a slightly contradictory impression.
None that I've found so far take condensed or expanded
fonts into explicit consideration when discussing this,
and often the exposition on this point is very short.


It appears, however, that an em should be the (nominal?) width
of a capital M.  This is also useful for horizontal measures:
em space, em dash, and the indent for the first line of paragraph
is best measured in em [as width of M, also for bold, expanded,
or compressed], as well as table column sizes that one wants
to tie to the character width.

The em, as width of M, is not very useful as a vertical measure.
There are three, as I see it, vertical relative measures a font
USER is interested in: the p height, the cap height, and the x
height. The typeface DESIGNER may well be interested in other
vertical and horizontal measures, like the descender to ascender
height, but that does not appear at all interesting for the font
USER.

It is especially important to get this done correctly for the
web, where for presentation one does not know the EXACT font(s)
used, but a font substitution is done.  Giving the size for
a square that is not visible to the ordinary viewer is not
a good idea, especially since that square (misnamed, in the
confusion, an "em square") has either no fixed relation to
the actual glyph sizes, or (Opentype?) focuses on a measure
(descender to ascender height) that is *NOT* of major interest
to the font USER.


Let me repeat what I've suggested before on www-style (I don't
think it got copied to www-font), with some added usage examples:

[Swashes, or swash-like features, don't count in the descriptions
below.]

Horizontal relative measures of interest to the user:

em -- width (still!) of an "M" in the current font; useful for
	specifying first line indent width, the widths of table
	columns (or even tab stops); NOT particularly useful for
	vertical measures.

tmu -- em/18; for convenient (horizontal mostly) adjustment in
	math expressions (note that math don't use excessively
	condensed or expanded typefaces, so tmu can there be used
	vertically too, without surprises).


Vertical relative measures of interest to the font user:

p-height -- height from (nominal) descender to top of first level
		accents on top of capital letters; the recommended
		minimum baseline distance; NOT particularly useful for
		horizontal measures.

H-height -- (nominal) cap height, useful for specifying consistent
		size, even under font substitution, for ideographic text
		as well as Latin/Greek/Cyrillic cap only headings.

x-height -- (nominal) x-height, useful for specifying the size of
		lowercase dominated text, especially when font substitution
		is done (count on that it's done, if published on the web).


Vertical relative measures **NOT** of interest to the font user
(in my opinion at least):

square size -- (misnamed: em square); no known relation to actual
		glyph size. (Why would any user be interested in this?)

descender to ascender -- (square for Opentype); does not take
		account of accents over capital letters (unless you
		have a special meaning for 'ascender' in Opentype).


Another repeat: As I have already suggested for CSS3, font-size
could set either the p-height, the cap height, or the ex height,
as given by a 'mode' parameter (as originally suggested by Ian
Hickson).  For best backwards compatibility, I've suggested
that the p-height is what is set by default.

Aside: setting the cap height to the inherited ex height is
useful for *fake* small caps.

Clarification: I see no point in directly setting the em width,
so I'm NOT suggesting any such mode for CSS3 "font-size".



		Kind regards
		/kent k




> -----Original Message-----
> From: Clive Bruton [mailto:clive@typonaut.demon.co.uk]
> Sent: Monday, January 31, 2000 11:08 PM
> To: www-font@w3.org
> Subject: Re: What's an em (was RE: Units, font sizing, and zoom
> suggestion for CSS 3)
> 
> 
> Karlsson Kent - keka wrote at 31/01/00 17:21
> 
> >	An em was originally the width of an M (swashes don't count).
> 
> Sorry, I think that is (mostly) just a coincidence, an em is 
> a square (ie 
> it has two dimensions, not one) and is defined by the *body* 
> size of a 
> piece of movable type. It is scalable since it relates 
> directly to the 
> size of type in question, an exception to this is that in 
> general people 
> measuring by "ems" mean "pica ems" which are 12pt ems (though this 
> doesn't apply in the CSS model).
> 
> >	In CSS an em is the "font-size", and relatively few use 
> it as yet.
> >		I haven't seen any good argument for keeping it "as is",
> >		other than an unwillingness to correct a mistake and/or
> >		to align with TeX. (Call it 'wem', or 'tem' if you like,
> >		but then deprecate the current 'em'.)  
> >
> >		Correcting the definition of an em to what it 1) usually
> >		means (width of an M), and 2) what it means 
> (width of an M)
> >		in *the* other *major* electronic typesetting 
> context (TeX)
> >		should have little impact for most.  If someone 
> asks for a
> >		line-height of 1em, then they should get what 
> they ask for,
> >		a line height the same as the width of an M, 
> nothing else.
> 
> This makes no sense. In your argument the line height would have no 
> relationship whatsoever to the size of the type. Conceivably 
> if one were 
> setting line height for a condensed face it would have to be set in 
> excess of 2.0em, on the other hand for an expanded face it 
> could be as 
> little as 0.5em. In both cases the perceived line height would be the 
> same.
> 
> And how would you specify line-height when you mixed different width 
> fonts on a single line, this would happen even if you were 
> setting common 
> combinations like bold and italic styles!?
> 
> The measurement of line-height or em size needs to be 
> explicitly linked 
> to a typeface's size, not its width.
> 
> 
> -- Clive
> 
Received on Tuesday, 1 February 2000 15:47:36 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 27 April 2009 13:54:03 GMT