Re: [css3-writing-modes] Request to publish Last Call Working Draft

Simon raises some excellent points!

Re non-textual entities:

> What cases are intended to have a different
> rendering than that which would result if this phrase were omitted, and
> the image was treated as a neutral character preceded by LRE/RLE
> /LRO/RLO and followed by PDF?

The two are definitely different. For <span dir=ltr>man<img dir=rtl
.../>dog</span>, treating the image as an R or AL character would put "man"
to the right of it and "dog" to the left, while treating it as being
surrounded with RLE and PDF would reverse the order (in the absence of
spaces or other neutral characters around the <img>).

As far as I can see, however, IE8, FF 4, and Safari 5.0.3 all treat the
<img> above neither as R nor as RLE PDF, but as just a neutral character,
plain and simple. (Can't test Chrome right now, but I presume it's the
same.) I say this because in the three first divs below, the "man" comes out
on the left in all three browsers, which is not the case for the last two
divs, which use a span instead of an img:

<!DOCTYPE HTML><html><title><head>img-with-dir</title></head>
<body>
<div dir=rtl>man<img dir=rtl src="x.png">dog</div>
<div dir=rtl>man <img dir=rtl src="x.png"> dog</div>
<div dir=rtl>man <img style="unicode-bidi:embed;direction:rtl" src="x.png">
dog</div>
<hr>
<div dir=rtl>man<span dir=rtl>.</span>dog</div>
<div dir=rtl>man <span dir=rtl>.</span> dog</div>
</body></html>

Is there any strong reason to specify this when no browser implements it?

My personal preference is that non-textual entities be treated as a neutral
character always. People sometimes use dir on them in order to affect the
direction of their alt or title, and I do not think that anyone expects them
this to affect the bidi ordering outside them.

> If 'direction' has the value 'rtl', should the image be considered as
> having character type R or AL? Consider the case of an image in Arabic
> text followed by a mathematical formula.

Another reason to prefer neutrality :-)

I would like to add a comment of my own on this topic: unicode-bidi:isolate
most certainly should *not* make the element behave as a strong character
(or as RLE/LRE/RLO/RLO/PDF). The current wording unfortunately implies that
it should.

Re the value syntax for unicode-bidi, the issues Simon raised seem to be
addressed in the latest draft: "normal | embed | [ isolate || [ plaintext |
bidi-override ] ]"

Aharon

Received on Thursday, 26 May 2011 06:19:00 UTC