Re: "phonemes" property in the CSS3 Speech module

Thank you for the illustration Tab.
For what it's worth, the current draft of EPUB3 reuses the element  
semantics from SSML in attribute form ("ssml:ph" and "ssml:alphabet"):

http://epub-revision.googlecode.com/svn/trunk/build/spec/epub30-overview.html#sec-tts

Regards, Daniel

On 4 Feb 2011, at 00:28, Tab Atkins Jr. wrote:

> Just to distill down the essential problem you have, and provide a
> tl;dr version:
>
> Authors may start with pages like this:
>
> <style>
> .a1678 { /* stupid class names are unfortunately common */
>  phonemes "toe-MAH-toe";
>  font-weight: bold;
> }
> </style>
> <p>My doctor said to eat a <span class=a1678>tomato</span> every  
> day.</p>
>
> And then, at some point in the future, it gets changed to:
>
> <p>My doctor said to take my <span class=a1678>vitamins</span> every  
> day.</p>
>
> (With the <span> being cargo-culted in because of the visual styling.)
>
> Now, screen readers will say "My doctor said to take my toe-MAH-toe
> every day.", to nonsensical results.
>
> The problem here is the indirection for what is really a property of
> the content.  You instead propose to do something like:
>
> <p>My doctor said to eat a <span
> pronounceas="toe-MAH-toe">tomato</span> every day.</p>
>
> Then, if the content changes in the future, it's much more obvious
> that this is wrong:
>
> <p>My doctors said to take my <span
> pronounceas="toe-MAH-toe">vitamins</span> every day.</p>
>
> ~TJ
>

Daniel Weck
daniel.weck@gmail.com