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

Re: Pixel densities (was RE: Best practice for font control)

From: Jonny Axelsson <jonny@metastasis.net>
Date: Wed, 26 Jan 2000 13:30:08 +0100
Message-Id: <>
To: www-style <www-style@w3.org>
I have always been fascinated with the "relatively absolute" pixel length
unit, both
(*) pixels as implemented
(*) pixels as defined (in CSS)

Onscreen pixels-as-implemented is unproblematic. The absolute size of a
pixel varies widely with screen resolution and pixel density. That is as it
should be, it makes it possible to render an HTML page on most displays and
get the relative proportions right. In addition it is a good fallback to
UAs that are not CSS-enabled, as HTML units are either pixels or fractions
(* and %). Included elements, like images, are practically always measured
in pixels.

Pixels as printout works well too, but there is a problem in that the
conversion to absolute length used in printouts use different pixel
densities. A Mac would print 72 pixels to the inch, making a "print pixel"
exactly like a point (which is by current definition precisely 25.4/72 mm),
a Windows PC 96 pixels to an inch. An image or a column defined to print
out snugly on an A4 page (or 8.5"x11") for a PC, that is anything about 750
px width, will be clipped on a Mac, which can handle little more than 600 px.

The CSS pixels-as-defined has two parts. One, a pixel density of 90px/inch
(a length unit of 0.2822222222 mm for us in the SI camp). Two, this is
measured "at arm's length". I presume this was to make this valuable unit
extensible to other media than computer screen and printout. You could say
that 4px would be borderline readable for people with normal vision on
quality display, no matter the medium (7-8 px, using special fonts and
VERSALS is bordeline readable on a computer screen, but that is no quality
display). From that you could calculate how large the letters would have to
be to make an advertisement on the surface of the moon readable for us. A
pixel from any defined distance is an absolute unit.

Pixels-as-an-angle is intriguing. An image 550px wide would take up half
this screen, *no matter the resolution* (whether 480x640 or 1200x1600). The
only way I could make it appear smaller would be to walk away from the
screen, and if the definition is to be taken literally, the image should
grow to compensate. Pixels-as-defined is the perfect unit for retinal laser
displays... But should a CSS compliant printer have a zoom option keyed to
expected view distance? If it is a poster (at average 7 m viewing
distance), print the page out at 1000%. I have seen the definition as more
heuristic than absolute, since how-far-is-the-user-from-me would be one of
the variables hardest for a system to judge. It would imply, however,
things like "a TV pixel should be 2-4 mm" (should it?), but when it faxes
the screendump, the pixel should be 0.28 mm.

Received on Wednesday, 26 January 2000 07:31:34 UTC

This archive was generated by hypermail 2.3.1 : Monday, 2 May 2016 14:26:52 UTC