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

RE: What's an em

From: Clive Bruton <clive@typonaut.demon.co.uk>
Date: Wed, 2 Feb 2000 12:22:16 -0500 (EST)
Message-Id: <S.0000076074@mail.indx.co.uk>
To: <www-font@w3.org>
cc: <www-style@w3.org>
Karlsson Kent - keka wrote at 02/02/00 16:20

>Assume the (lowercase dominated) headings are asked (by the page author)
>to be of x-height 7 mm in Flemish Script, the body text (lowercase
>dominated) to be of x-height 3.5 mm Verdana.

I think trying to specify type on screen in mm would be a mistake, but...

>Whatever font substitutions
>are done (keeping the asked-for x-heights to a reasonable degree), one
>would still get a proper percieved size difference between the heading
>and the body text.  Right? Or???

Yes, great. What would the line-height be? For a clue on this look at the 
examples on the CSS2 page, where line-height is maintained and ascender 
and descenders clash where the x-height has been taken as the defintive 

>An invariant you would NOT get if the
>heading and body sizes were asked for in "point sizes" for some unknown
>square.  Right?  'Definiteveness' is a non-issue when the text is hard
>to read, or even unreadable.

No, you aren't going to have 12pt Flemish script switched for 12pt 
Verdana, it's going to be 18pt Verdana substituted for 18pt Flemish 
script - so the difficulty of reading text isn't the issue, the scale of 
text to headings is.

As noted above, if you scale based on x-height you end up with 
acsender/descender clash.

>Please read my suggestion again before entangling everyone in strange
>calculations that have nothing to do with the suggestion at hand.

Strange calculations that show what happens in extremes, or in real world 

Please, show me a "real world" calculation that fits your model?

>But as I said before, my issue is not with extreme typefaces, where 
>some odd (but hopefully not too odd) effects may occur, but with
>ordinary "every-day" typefaces.  So even if it is not perfect, *some*
>adjustment to get *near* size invariance is better than complete
>non-invariance.  Which does give problems NOW, when different systems
>do different font substitutions, and the author might have no idea
>what font substitutions are done (and would not care to test them all,
>even if he/she did).

If you are trying to fake perceived size you need to do it properly based 
on x, cap, ascender and descender factors, plus character widths and stem 
weights. Trying to isolate one of those factors just leads to erroneous 
output, but the only one that is going to get anywhere near close is pt 
or em size.

If you scale x-height you also have to scale the other factors of 

-- Clive
Received on Thursday, 3 February 2000 10:21:28 UTC

This archive was generated by hypermail 2.3.1 : Monday, 2 May 2016 14:26:53 UTC