Re: problems with Arabic joining rules

From: Asmus Freytag (asmusf@ix.netcom.com)
Date: Mon Jun 28 1999 - 19:37:18 EDT


Werner,

Good catch. Rules R3 , R5 and R6 are equally broken.

It appears as if the words "left" and "right" were left out of the text of
these rules in a few places.

Below table 6.3 two 'superset classes' were defined, to wit:

right-join causing

        and

left-join causing

The fact that right join-causing is then never referred to is a good
tip-off as to the source of the error. Here's the fix:

In all but rule R2, the words left or right need to be inserted in front of
the words join-causing.

R3: join causing -> left-join causing
R4: join causing -> right-join causing (first instance)
     join causing -> left-join causing (second instance)
R5: join causing -> right-join causing (first instance)
     join causing -> left-join causing (second instance)
R6: join causing -> left-join causing (first instance)
     join causing -> right-join causing (second instance)

It would probably also help if the two superset classes were to be added
into table 6-3 instead of being defined in the text, since the current
arrangement makes the table 6-3 incomplete.

Table 6-3 would then contain these additional lines below 'join causing'

right-join causing | superset of dual-joining, right-joining and join-causing
left-join causing | superset of dual-joining, left-joining and join-causing

The first sentence following table 6-3 can then be deleted.

I have forwarded this as an erratum so it can be fixed in Version 3.0
A./

At 01:20 PM 6/28/99 -0700, Werner LEMBERG wrote:
>
>Dear friends,
>
>I have problems with the Arabic joining rules given on page 6-25f in
>the Unicode 2.0 book
>
>Example:
>
> R4 says that `a dual-joining character X that has a join-causing
> character on the right and a join-causing character on the left will
> adopt the [medial] form X_m'.
>
>In a previous paragraph, the terms `dual-joining' and `join-causing'
>are well defined, both defining classes of glyphs. But the class
>`join-causing' consists basically only of TATWEEL! With other words,
>R4 says that the medial form of a character appears only if on both
>the left and right side is a TATWEEL character.
>
>But this can't be true -- either I have to insert TATWEEL characters
>implicitly (which isn't mentioned anywhere in the description of the
>algorithm), or the term `join-causing' must be replaced with `right
>join-causing' resp. `left join-causing'. This affects other rules
>also.
>
>My knowledge of Arabic is too limited to verify this. In case it is
>an error, please give me a pointer where to find an updated version of
>the algorithm (and mention that fact on the Errata home page). In
>case I've misunderstood something, please explain how to proceed
>correctly.
>
>
> Werner
>
>



This archive was generated by hypermail 2.1.2 : Tue Jul 10 2001 - 17:20:47 EDT