Support for CSS selectors and positions in textref

[HadrienGardeur]

Now that the Guided Navigation syntax is also being considered for read aloud (see #4), we should also consider adding support for two other fragment types on textref IMO:

• CSS selectors using #css()
• and/or positions using #position=X

When working with text, these additional fragments are not necessarily meant for locating text fragment precisely but can help to disambiguate between two similar text fragments.

To support a read aloud feature, a reading application would generate a Guided Navigation Document where text contains the text that will be passed to the TTS engine (although it will most likely be pre-processed first, for example to break it down into utterances) and textref would reference the source HTML file and the fragment that should be highlighted as the text is being read.

This would also open the door to textual/audio alignment after an ebook has been authored, for example automatic alignment between an audiobook and an ebook.

[mickael-menu]

I agree that #css() is a must and will be very useful.

For #position=x, I'm not sure we can really use it as we don't have accurate position information when processing the HTML content. As the Readium position is computed from the raw bytes length and not character count.