Re: Another --> Re: GtkMovementStep of GtkTextView

[Pablo Saratxaga <pablo mandrakesoft com>]

Kaixo!

On Mon, May 21, 2001 at 09:27:19AM -0700, Chookij Vanatham wrote:
 
>> Those have also their own keys on Arabic keyboards I think.
> 
> So, I think that, in arabic script, users like to have <backspac> to delete
> the previous key typed. This is the same as Thai/Indic/Korean/Vietnamese...

No, not in Vietnamese.
In Vietnamese (in telex input mode) the last typed key can be a 'j' or a 'w'
but the last visible letter can be an 'n' or a 'g' or...

I think the important thing is if the user can make a link between what
has been typed and what shows on screen.

The user can do the link if he has just typed it, even if it is very 
different than what shows on screen; that is why I think it can make sense
to have two behaviours, one for backspacing just after typing something
(instant correction), where the effect should be to undo last keystroke;
and another behaviour when editing an already existing text, where the
behaviour should be linked to what is seen on screen, and not to what
is shown.

Note also that in Korean you type either in latin romanization, or in
korean alphabet (not korean syllabs, but korean letters); however, what
is stored, are korean syllabs; so, there is no direct link between
what is typed and what is stored.
On the other hand, the structure of Korean syllabs make it possible to
see without any ambiguity its components, so having the BakSpace deleting
the components rather than the whole syllabe may be a choice.

> One difference in arabic is that those base letter with diacritic
> (U+0622..U+0626) will be treated as a single letter and that is also typed
> by one key stroke. The <backspace> algorithm still works with those letters
> because they have their own codepoints.

The typical user won't know about that, and won't care at all.
having a unicode code point is not argument (a lot of chars can also exist
as precomposed or can be encoded as base char and composing diacritics;
both representations on a text file should be handled the same way
when editing (unless, maybe, in special editing mode where all special
chars are shown)

> <backspace> algorithm is just removing
> one codepoint from logical stream...

That is indeed the most simple, and will work on most cases.
However, I don't think the user should handle differently a same visual
"udiaeresis" if it is one single unicode char, or 'u' + 'combining diaeresis'
For latin, greek and cyrillic, the preferred behaviour would be
to remove the whole char at once.

But, for Thai you say the preferred behaviour is the opposite.

So, the two are needed, and the behaviour would change depending on the
script, and/or user configuration.

-- 
Ki ça vos våye bén,
Pablo Saratxaga

http://www.srtxg.easynet.be/		PGP Key available, key ID: 0x8F0E4975