W3C home > Mailing lists > Public > www-svg@w3.org > November 2002

Re: Conformance Criteria - Device Pixel

From: Chris Lilley <chris@w3.org>
Date: Wed, 13 Nov 2002 09:25:23 +0100
Message-ID: <9452579000.20021113092523@w3.org>
To: www-svg@w3.org, Jim Ley <jim@jibbering.com>

Hello Jim, www-svg,

  In this message
  you wrote:

> "All visual rendering must be accurate to within one device pixel to the
> mathematically correct result."
> I'm concerned this seems somewhat an excessive requirement, and with even
> quite reasonable numbers (large viewbox, small stroke width, complicated
> bezier path, high resolution output device,) we end up with conforming
> SVG viewers needing better than IEEE-754 Doubles precision calculations.
> Why the requirement on device pixels, rather than reference pixels?

This was discussed in the working group an in short, we agree.

Reference px units rather than device pixels makes more sense,
particularly given the possibility of (say) a 2400 dpi printer that
uses screening to simulatte continuous tones. Each dot is incredibly
tiny, and each dot can only be white, black, cyan, magenta or yellow
and is thus, very far from the correct color. Spatial resolution is
sactrificed to give color precision over an area roughly equivalent to
one px unit.

We also discussed the effect of zoom levels (for example, zooming way
in on a curve, thus exposing the precise details of curve flattening
or tesselation) and the group considers that visual rendering to one
reference pixel refers only to the initial zoom level. Clearly viewers
should attempt to high degree of accuracy when zooming but as you note
this cannot be done in the general case without infinite precision
arithmetic and thus should not be part of the conformance criteria as

 Chris                          mailto:chris@w3.org
Received on Wednesday, 13 November 2002 03:25:23 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 8 March 2017 09:46:54 UTC