From: Chris Lilley <chris@w3.org>
Date: Sun, 21 Aug 2005 01:59:57 +0200
Message-ID: <64743907.20050821015957@w3.org>
To: "Doug Schepers" <doug@schepers.cc>

```
Hi Doug, good to meet you last week.

Thanks for the thoughtful discussions on an auto direction algorithm.
This email is just closing the loop and directing you to what the Tiny

On Saturday, November 13, 2004, 10:45:41 AM, Doug wrote:

DS> While there is a certain lack of precision, there is an equal lack of
DS> commitment; if a person is slightly surprised by an unorthodox navigation
DS> decision, they can quickly figure out the system behind it and navigate to
DS> the correct option, and the can't get "stuck"; with document order, they
DS> have no cues at all to try to recover from a mistaken direction navigation.

This is helpful, and we have considered various automatic algorithms
such as Thiessen polygons derived from shape centroids, however its
clear that no one algorithm copes well with all content (consider
intersecting shapes, shapes with 'donut holes' which contain and
completely surround other shapes, etc). Its also become clear that the
user can indeed get 'stuck' (moving back and forward skips over a
particular shape of interest that cannot be reached) and that
unintuitive results can often be produced (going in one direction and
then going in the reverse direction frequently does not take you back to
where you started, in many algorithms).

DS> There are other, more complicated approaches, like using a Voronoi diagram
DS> or Delaunay triangulation, but in this case, I think that the Good-Enough
DS> principle holds. There are drawbacks to every approach I could think of,

Right. We found that too.

DS> and the one I went with finally was the simplest to do, and equally
DS> intuitive to the end user. The point is, any directional solution is
DS> better than a purely "hidden" document order selection.

Given that, and given the difficulty of finding an algorithm that is
both general and gives intuitive and reproducible results, and given
that things like the accuracy of tessellating curves will result in
different results on different implementations - always a bad thing) the
WG decided as follows.

Firstly, to add a way for the document author to add specific
directional navigation (linking to a given shape by ID, so they can
design a system that is intuitive)

and secondly, using document order as a fallback mechanism since it
gives repeatable results. we agree that it is not a preferred mechanism,
here its just a fallback so that *something* happens rather than focus
staying on the same element.

--
Chris Lilley                    mailto:chris@w3.org
Chair, W3C SVG Working Group