Re: [SVGMobile12] Lack of BIDI 'direction'

Hello www-svg,

Fantasia wrote:
> There's discussion in about
> bidirectional text, but there is no definition for a 'direction' property
> that I can find, nor any mention of any substitute for it (such as a 'dir'
> attribute).

That is correct. Thank you for pointing out this omission. SVG 1.1 has
a direction property, as does SVG Basic 1.1. SVG Tiny 1.1 did not have
it, because the text element could not have a tspan child element in
Tiny 1.1 and thus adittional levels of embedding were not possible.

You are correct to point out that SVG Tiny 1.2 should have the
direction property restored.

> This is a problem because the Unicode bidirectional algorithm needs a
> base direction for each "paragraph" that it operates on. For plain text
> it guesses based on the first character in the text, but this is not
> always accurate.

Yes, if the initial character is directionally weak, or neutral, it
will not give the expected results and must be overidden.

> Applying this rule to SVG Tiny would be incompatible
> with an implementation that supported SVG Tiny plus the 'direction'
> property because 'direction' defaults to 'ltr', not to 'guess a direction'.

However, recall that the Unicode bidi algorithm still applies, unless
the 'unicode-bidi' property is set to 'embed' or 'bidi-override'. The
initial value is 'normal'. Thus, unless 'unicode-bidi' is set to a
non-default value, implementations which support the 'direction'
property do in fact use the Unicode bidi algorithm to determin

But this simply means that we need to restore the 'unicode-bidi'
property from SVG 1.1 (and Basic 1.1) as well.

> The lack of a 'direction' affects both bidi reordering and other properties
> like 'text-align', whose values' behavior depends on 'direction'.


> I therefore strongly recommend that SVG Tiny include the 'direction'
> property.

The SVG WG has discussed your recommendation and agrees with it. In
response you your last call feedback, SVG Tiny 1.2 will re-instate
both the 'direction' and the 'unicode-bidi' properties from SVG 1.1.

An editors draft of the specification, incorporating this change, is
available at

The SVG WG has also added a test for these two properties, and has
determined that there are at least two passing implementations
(because these properties were previously in the SVG specification).

Thanks once again for your bringing this deficiency to our attention.
We hope that this satisfies your concern; please let us know.

 Chris Lilley          
 Technical Director, Interaction Domain
 W3C Graphics Activity Lead
 Co-Chair, W3C Hypertext CG

Received on Friday, 10 October 2008 19:12:44 UTC