RE: New specdev material about bidi

   Hello, Richard!

On August 3, you wrote:
<quote>
> 1) In section 5.2.4, we find: "They are invisible in most editors and are therefore difficult to work with, and can easily lead to orphans and overlapping ranges."
> I think this is not strictly correct. Since each CC such as LRE/RLE or LRI/RLI starts a new embedding level, there cannot be overlapping ranges other than total inclusion, which in itself is not an error.

But you could end up with "text LRI more text RLE more text PDI more text PDF".  And in HTML, because of a clause that the editor put in which says that implicit closing of LRI/RLI by PDF is not allowed, you could also end up with the inverse.  One additional way this overlap could occur would be if part of the logic is in markup and part in stray CCs. (see https://github.com/whatwg/html/issues/709)

Do you agree?
<end of quote>

I don't agree. The explanations of rule X6a of the UBA say, in quite convoluted terms, that a PDI terminates not only the isolate sequence started with the last FSI/LRI/RLI but also all open embedding sequences within that isolate sequence.
In your example, the PDI would close the embedding level started by the RLE, so that the trailing PDF would become an orphan. So, no, there cannot be overlapping ranges other than total inclusion, IMHO.
Of course, the case in the example is ill-formed and the result will probably not be what the author intended.

Mixing markup and CCs is not recommended, but when that happens, good implementations must translate both to some common representation, so that the UBA rules can be applied uniformly to the mixture. Thus it is possible to create text with unmatched embedding and isolating initiators and terminators, but we will never have overlapping ranges.
So it seems to me.

Shalom (Regards),  Mati

Received on Sunday, 7 August 2016 23:23:32 UTC