RE: Mirroring Unicode symbols in Arabian

Certainly if mirrored glyphs have different semantics, i.e., might appear in the same document together and not because of changing math-zone directionality, then they should be encoded as characters. This is the case for bracket pairs, line drawing characters, arrows and related symbols. But it’s not the case for many characters, such as integrals and summations. It’s better for such characters to have a single code point, independent of the math-zone directionality. It clearly simplifies math search and content considerations not to have to worry about BiDi directionality.

Murray

From: Paul Topping [mailto:pault@dessci.com]
Sent: Wednesday, February 20, 2013 2:31 PM
To: Neil Soiffer; Murray Sargent
Cc: Tom Leathrum; Kent Karlsson; David Carlisle; Daniel Marques; www-math@w3.org; Khaled Hosny
Subject: RE: Mirroring Unicode symbols in Arabian

Neil,

I agree with your points but they assume that the use of the mirrored glyphs is only for RTL. In other words, the mirrored character does not have different semantics from the one it mirrors. Since mathematics does use mirrored glyphs to represent alternate, but usually related, semantics, I felt it must be pointed out that such uses are not covered by mechanisms being discussed here. Should any mathematician on this list be thinking that they’d like to be able to use mirrored characters to express their mathematical concepts, they might want to know think again.

Paul

From: neil.soiffer@gmail.com [mailto:neil.soiffer@gmail.com] On Behalf Of Neil Soiffer
Sent: Wednesday, February 20, 2013 2:22 PM
To: Murray Sargent
Cc: Paul Topping; Tom Leathrum; Kent Karlsson; David Carlisle; Daniel Marques; www-math@w3.org; Khaled Hosny
Subject: Re: Mirroring Unicode symbols in Arabian

The mirrored renderings are just glpyhs and are not treated by Unicode any differently than a serifed, bold, or large version of a given character.  A PUA would only be used if you had a different *character*, not a different *glyph* for an already defined code point.

What needs to happen is for font vendors to develop fonts with mirrored images in them for people who want to use math in an RTL context.  MathML has a "dir" attribute and should pick up the default direction from the environment, so a renderer is able to switch to a different font or pass that info along to a font call just as they can if they choose to match a other aspects of the math rendering with the environment in which they are displaying.

     Neil


On Wed, Feb 20, 2013 at 2:07 PM, Murray Sargent <murrays@exchange.microsoft.com<mailto:murrays@exchange.microsoft.com>> wrote:
I really do not like the idea of using PUA code points for mirrored glyphs except perhaps as a private hack. PUA codepoints are nonstandard by definition and should be limited to experimental usage. (I've been a member of the Unicode Technical Committee since 1994, so perhaps I'm a bit brainwashed :-)) Meanwhile the OpenType approach is a standard, widely used way of doing such things.

Murray

-----Original Message-----
From: Paul Topping [mailto:pault@dessci.com<mailto:pault@dessci.com>]
Sent: Wednesday, February 20, 2013 1:13 PM
To: Tom Leathrum
Cc: Neil Soiffer; Kent Karlsson; David Carlisle; Daniel Marques; www-math@w3.org<mailto:www-math@w3.org>; Murray Sargent; Khaled Hosny
Subject: RE: Mirroring Unicode symbols in Arabian

Right. I just wanted to point out to the math buffs on the thread that if the only mechanism to show reversed glyphs is via RTL or some app-specific access to OpenType features. I am not sure if CSS gives the necessary control to show these glyphs in HTML content. If anyone wants these reversed glyphs to be used in math, then font vendors should be encouraged to give them PUA points in their fonts and, presumably, there would be a case to be made to propose non-PUA points for them in Unicode. My point may be off-topic.

Paul

> -----Original Message-----
> From: Tom Leathrum [mailto:leathrum@jsu.edu<mailto:leathrum@jsu.edu>]
> Sent: Wednesday, February 20, 2013 12:10 PM
> To: Paul Topping
> Cc: Neil Soiffer; Kent Karlsson; David Carlisle; Daniel Marques; www-
> math@w3.org<mailto:math@w3.org>; Murray Sargent; Khaled Hosny
> Subject: Re: Mirroring Unicode symbols in Arabian
>
> If I may interject here:
>
> Paul, I think Murray's point about semantics addresses your concern,
> because if there is a semantic difference between reversed symbols in
> LTR then they will be represented in different Unicode values.
> Consider for example U+2282 "subset" and U+2283 "superset" -- the
> glyphs are mirror-images, and both are marked as mirror="Y" for RTL.
> RTL would have the visual effect of swapping these two glyphs, but in
> fact their Unicode values wouldn't change because the semantics would
> still be the same -- U+2282 would be semantically "subset" whether it
> is mirrored or not, and in RTL the glyph for it would look (to LTR
> readers) like a superset symbol.  Clear as mud?
>
>
> ----- Original Message -----
> From: "Paul Topping" <pault@dessci.com<mailto:pault@dessci.com>>
> To: "Murray Sargent" <murrays@exchange.microsoft.com<mailto:murrays@exchange.microsoft.com>>, "Khaled Hosny"
> <khaledhosny@eglug.org<mailto:khaledhosny@eglug.org>>
> Cc: "Neil Soiffer" <neils@dessci.com<mailto:neils@dessci.com>>, "Kent Karlsson"
> <kent.karlsson14@telia.com<mailto:kent.karlsson14@telia.com>>, "David Carlisle" <davidc@nag.co.uk<mailto:davidc@nag.co.uk>>,
> "Daniel Marques" <dani@wiris.com<mailto:dani@wiris.com>>, www-math@w3.org<mailto:www-math@w3.org>
> Sent: Wednesday, February 20, 2013 1:46:15 PM
> Subject: RE: Mirroring Unicode symbols in Arabian
>
> So this means that reversing a symbol to express a mathematical
> concept is unavailable to non-RTL text even though the fonts and the
> font rendering mechanism has the capability.
>
> Paul
>
> > -----Original Message-----
> > From: Murray Sargent [mailto:murrays@exchange.microsoft.com<mailto:murrays@exchange.microsoft.com>]
> > Sent: Wednesday, February 20, 2013 11:36 AM
> > To: Paul Topping; Khaled Hosny
> > Cc: Neil Soiffer; Kent Karlsson; David Carlisle; Daniel Marques;
> > www- math@w3.org<mailto:math@w3.org>
> > Subject: RE: Mirroring Unicode symbols in Arabian
> >
> > I meant that the codes for the mirrored integral, etc., are exactly
> the
> > same as the unmirrored symbols. The display software just mirrors
> them
> > in RTL math zones. If you use OpenType, you used the 'rtlm' feature
> or
> > shaping, as the characters warrant. This is the same thing as for
> > mirrored characters that have mirrored character counterparts. An
> open
> > paren is U+0028 whether it's mirrored or not.
> >
> > Murray
> >
> > -----Original Message-----
> > From: Paul Topping [mailto:pault@dessci.com<mailto:pault@dessci.com>]
> > Sent: Wednesday, February 20, 2013 11:31 AM
> > To: Murray Sargent; Khaled Hosny
> > Cc: Neil Soiffer; Kent Karlsson; David Carlisle; Daniel Marques;
> > www- math@w3.org<mailto:math@w3.org>
> > Subject: RE: Mirroring Unicode symbols in Arabian
> >
> > Not sure how this answers my question.
> >
> > Paul
> >
> > > -----Original Message-----
> > > From: Murray Sargent [mailto:murrays@exchange.microsoft.com<mailto:murrays@exchange.microsoft.com>]
> > > Sent: Wednesday, February 20, 2013 11:24 AM
> > > To: Paul Topping; Khaled Hosny
> > > Cc: Neil Soiffer; Kent Karlsson; David Carlisle; Daniel Marques;
> www-
> > > math@w3.org<mailto:math@w3.org>
> > > Subject: RE: Mirroring Unicode symbols in Arabian
> > >
> > > Mirrored glyphs are a display feature, not a semantic one.
> > >
> > > Murray
> > >
> > > Paul Topping asked, "If access to these characters requires use of
> > > this OpenType feature, does it imply that such characters will not
> be
> > > accessible from applications that simply process Unicode text
> strings
> > > (eg, web browsers and most other apps)?"