Re: Proposal for isolation characters in Unicode and the unicode-bidi:isolate and unicode-bidi:plaintext definitions

I am preparing a new version of the proposal, but I would like to mention
separately that, as Martin suggested, it will include a new character to
terminate isolates, which I am calling PDI (POP DIRECTIONAL ISOLATE).

The question naturally arises how the algorithm should behave when isolates
are not properly nested with respect to embeddings/overrides, e.g.
a: RLI LRE PDI PDF, and
b: RLE LRI PDF PDI.

There are basically two possibilities:

1. Make isolates weaker than embeddings/overrides, i.e. ignore a PDI when a
PDF is expected, and have a PDF close all isolates opened between it and
its marching LRE/RLE/LRO/RLO. Thus, in a, the PDI is ignored, and in b, the
PDF ends the scope of the RLI as well as the LRE.

2. Vice-versa - make isolates stronger than embeddings/overrides, i.e.
ignore a PDF when a PDI is expected, and have a PDI close all
embeddings/overrides opened between it and its marching FSI/LRI/RLI.
Thus, in a, the PDI ends the scope of the RLE as well as the LRI, and in b,
the PDF is ignored.

Possibility 2 offers greater forward compatibility, since new and old apps
will interpret the PDFs as closing the same scopes when isolates are not
properly nested with respect to embeddings/overrides.

Possibility 1, on the other hand, gives isolates the desirable feature of
isolating their surroundings from their contents - even when their contents
contains extra or missing PDFs.

I have decided to go with possibility 2, since IMO forward compatibility is
not very important for what are, essentially, broken documents.

Aharon



On Tue, May 15, 2012 at 11:10 AM, "Martin J. Dürst"
<duerst@it.aoyama.ac.jp>wrote:

> It may be worth considering to create a new character to close these
> embeddings. Otherwise, older algorithms will close LRE/RLE/LRO/RLO
> embeddings/overrides prematurely.
>