- From: r12a <ishida@w3.org>
- Date: Thu, 15 Sep 2016 11:15:00 +0100
- To: Simon Montagu <smontagu@smontagu.org>, Martin J. Dürst <duerst@it.aoyama.ac.jp>, "public-i18n-bidi@w3.org" <public-i18n-bidi@w3.org>
- Cc: Roozbeh Pournader <roozbeh@google.com>, "Aharon (Vladimir) Lanin" <aharon@google.com>, Shervin Afshar <shervinafshar@gmail.com>, Mostafa Hajizadeh <mostafa@daftar.cc>
On 15/09/2016 10:49, Simon Montagu wrote: > On 15/09/16 07:51, r12a wrote: >> On 15/09/2016 05:44, Martin J. Dürst wrote: >>> This is a very high level, speculative comment, but I'll make it anyway: >>> >>> You sound as if the isolates are too isolated. My understanding is that >>> we introduced the isolates because the embeddings were not independent >>> (isolated) enough and interacted with their surroundings too much. >>> >>> Did we overdo (if maybe even just so sligthly) the isolation when we >>> created isolates? Or would we (at least in theory) need a third kind of >>> range, somewhere in between isolates and embeddings in independency? >> >> i don't think the level of isolation is the problem, i think it's more >> to do with an isolated range being treated as a neutral character >> (whereas a non-isolated embedded range (eg. RLE) is treated as a strong >> character). >> >> ri >> > > That sounds to me like the same issue: as soon as an embedded sequence > is treated as a strong character, it stops being isolated: for example > it can affect the resolved level of an adjacent numeral. IIUARC this was > one of the chief reasons, if not THE reason, for treating isolated > sequences as neutral characters in their containers i agree that it's probably an inseparable issue. The question is how to ascertain that a string like "RLI فعالیت بینالمللیسازی، PDI", which i think should be regarded by default as a RTL string can be perceived as such - especially if those controls have been added by something else along the way, such as an application that wraps strings, and which therefore removes the previously existing clues. Asmus, i hear what you're saying about higher level protocols, but i can't help thinking that those protocols would need to be adopted by just about any application that deals with strings of this kind - which makes me think that perhaps there should be a standard mechanism described by the UBA (?). ri
Received on Thursday, 15 September 2016 10:15:22 UTC