Re: Bob G. Comments on BiDi in HTML/i18n draft

From: Martin J Duerst (
Date: Thu May 23 1996 - 12:39:23 EDT

Jonathan Rosenne wrote:
>>From: "Robert N. Goldrich" <bobg@ACCENTSOFT.COM>
>> 1. No BiDi layout should be performed on text marked with
>> the <PRE> tag.
>If this is intended to support existing Hebrew pages ("visual"), this
>is not what PRE was intended for. They could be identified by the charset
>(8859-8, rather than 8859-8i or 10646).

Identification by charset would be very inconvenient (apart from mixing up
different parts of the architecture that should care for different things):
It would be necessary to have a list of all existing and future visual "charset"s.

>Or a &rlo;/&pdf; could be added to each line.
Or the whole thing could be included in a single <BDO DIR=LTR> </BDO>
markup. In this respect, markup is very convenient because it is
not automatically terminated at line ends. This of course opens
the question of what exactly
some text <SPAN DIR=RTL>text text text
text text</SPAN> text text text

actually means. Note that again here we have a conflict between meta level
and base level. The definition that currently makes most sense to me is
to define this a equivalent to:

some text <SPAN DIR=RTL>text text text</SPAN>
<SPAN DIR=RTL>text text</SPAN> text text text

but this seems to be a very difficult issue.

>If PRE and it's friends XMP and LISTING are defined to exclude HTML formatting
>but not bidi processing, it is always possible to add bidi override codes to
>suppess bidi processing, but is they are defined to exclude bidi processing
>there is no way to activate it if one really wants just to use PRE the way it is
>used in English.

There seems to be an important difference between PRE on the one hand
and XMP/LISTING on the other.
PRE can contain quite some other tags (currently A, HR, BR, SPAN, and BDO,
of which only SPAN and BDO are due to i18n). So it rather makes sense to
define it as also including Bidi processing.

On the other hand, XMP/LISTING can't contain any markup, and the idea
here really seems to have been to not allow anything fancy, so that in
this case, it seems reasonable to define that Bidi is not applied at all.

Regards, Martin.

This archive was generated by hypermail 2.1.2 : Tue Jul 10 2001 - 17:20:31 EDT