W3C home > Mailing lists > Public > w3c-wai-ig@w3.org > April to June 2001

Re: CSS - font-size

From: <joeclark@joeclark.org>
Date: Tue, 03 Apr 2001 03:38:14 +0200
Message-Id: <200104030138.f331cEV04466@zidane.inria.fr>
To: w3c-wai-ig@w3.org

>pixels are not relative in the sense of being relative to the other 
>types of font. They are relative in the sense that they are not the 
>same size on different machines. So they aren't very helpful for a 
>given user, since it is hard to change the size of a pixel.

That is irrelevant.

It is, however, pretty easy to change the size of type used on a Web 
page in compliant browsers.

And anyway, the Zeldman and others have documented that the only size 
settings that work in the real world are none at all or pixels.


>I think the best technique is to use percentages, or em units. The 
>problem with smaller and larger is that some browsers define them as 
>a LOT smaller - to the point of being hard to read - typical use 
>cases don't go that far. But there is no intrinsic reason not to mix 
>those with percentages / ems.

Except that you will always force a fraction of the user's preferred 
size. 0.8 em is 80% of the type size. 80% is the same as 0.8 ex, or 
80% of the user's lineheight, which is almost always assigned by the 
browser (and will cause illegibility in Netscape 4 if you set it 

So if I specify 17pt type as my default, and you decide that 0.8 em 
is what I really meant after all, I get 17*80% or 14pt type. Did I 
ask for that, exactly?

And if I try to fix what you did by resetting my default to 20 px, I 
still only get 16! I don't recall anyone giving you that kind of 
power-- to flat-tax my font sizes, knocking off 20% no matter what I 

And what about old browsers that misinterpret what those settings are 
a percentage *of*?

Now, obviously, if you set the type size at 13px, that isn't what I 
wanted, either. But it is an absolute measure and will produce 
reliable results on a majority of machines. If you argue that 13px on 
a high-resolution screen is very small, well, so will everything else 
be, except maybe on Mac OS X. And we're talking about low-vision 
users; their screen resolution is effectively enlarged anyway. Even 
for nondisabled users who prefer a bigger font size, what are the 
odds that they're using a high-resolution monitor *and* an old 
browser that doesn't let them resize type?

Besides, Charles has conceded that <smaller> and <larger> and the 
other five defined sizes don't work in the real world. Why recommend 
something else that doesn't?

The issue of font sizes is crap all around, and nothing works 
reliably on all browsers and platforms, or even on popular browsers 
and platforms. What works least unreliably are pixels or nothing at 
all. At some point, we have to expect cooperation from low-vision 
people to upgrade to browsers that permit font resizing no matter how 
fonts are specified in the stylesheet. IE 5 on Macintosh; Opera; most 
later Explorers on Windows (with limitations); and iCab (which 
ignores stylesheets altogether) are already there. Apparently so is 
IE 6 on Windows, and Mozilla and Netscape 6 on everything save for 
Macintosh, where a conflict with Adobe Type Reunion or something has 
to be worked out.

- -- 
         Joe Clark | joeclark@joeclark.org
         Author, _Building Accessible Websites_
         (New Riders Publishing, October 2001)
         <http://joeclark.org/book/> | <http://joeclark.org/access/> 
Received on Monday, 2 April 2001 21:38:16 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 13 October 2015 16:21:12 UTC