- From: Xidorn Quan <quanxunzhen@gmail.com>
- Date: Sat, 15 Nov 2014 08:14:04 +1100
- To: fantasai <fantasai.lists@inkedblade.net>
- Cc: Koji Ishii <kojiishi@gmail.com>, kawabata taichi <kawabata.taichi@gmail.com>, "L. David Baron" <dbaron@dbaron.org>, "www-style@w3.org" <www-style@w3.org>
- Message-ID: <CAMdq699vtb1M2xUwP=p=OYSghNTriKoKhvMyGYMakz_wUogWuA@mail.gmail.com>
On Sat, Nov 15, 2014 at 3:54 AM, fantasai <fantasai.lists@inkedblade.net> wrote: > On 11/13/2014 01:39 AM, Xidorn Quan wrote: > >> >> On Thu, Nov 13, 2014 at 5:24 PM, Koji Ishii <kojiishi@gmail.com <mailto: >> kojiishi@gmail.com>> wrote: >> >> On Thu, Nov 13, 2014 at 5:53 AM, Xidorn Quan <quanxunzhen@gmail.com >> <mailto:quanxunzhen@gmail.com>> wrote: >> >> >> I have another suggestion. I found that in all use cases >>>> I had seen in JLREQ and specs, spanning is never directly >>>> connected with any previous separate-paired annotation. >>>> Is that make sense to only have span when an annotation >>>> is the only child of a <rtc>? I think that could significantly >>>> reduce the complexity on width calculation (which is the >>>> hardest part in my opinion) and line breaking. In addition, >>>> even if we drop spanning completely, we have to process >>>> this level of complexity to support ruby-merge anyway. >>>> >>> >>> I don't understand what you meant by "connected", >>> >> >> I meant, I found that spans do not immediately follow other >> annotations, so that use cases for spanning can be covered >> by the solution I proposed. >> >> but do you mean to allow spanning only when there is only >>> >> > one <rt> child for a <rtc>? If that's the case, I think it's >> >>> reasonable. If I misunderstood what you meant, can you >>> clarify a bit more? >>> >> >> Yes, that's what I meant. That could significantly simplify >> handling spanning, since there won't be spans of different >> width in one segment. >> > > Yes, I don't see a problem here either. I would prefer, if it's > possible, that we only span if the content is directly contained > in an <rtc> rather than special-casing <rt>s that are the only > child. Is that workable? That was the original goal: to make > content directly contained by an <rtc> span all the bases. > The effects on <rt> was just error-handling that fell out of > that approach. I think it's workable. But I prefer the spec to say something like "If the only child of an <rtc> is an anonymous <rt>, which means that all content is directly contained by the <rtc>, it spans all the bases." so that we do not need to complexify the anonymous box generation part, and we won't apply something on <rt>s but forgot the content directly inside <rtc>s. Does that make sense? - Xidorn
Received on Friday, 14 November 2014 21:15:13 UTC