Editing Sinhala and Similar Scripts

[Doug Ewell]

"Whistler, Ken" <ken dot whistler at sap dot com> wrote:

> But what Peter was pointing out that in the *generic* case
> for editing, such as first cursor down at some random
> location in already existing text, there is no existing history
> of how that text was created. And thus there are no "keystrokes"
> to be undone by hitting a backspace at that point.

Well, of course you and Peter are right, and stated it well as always.

Probably a better way for me to say it would be that, for any visual
combination of letters or marks that are decomposable in some way, such
as an acute accent over an 'e' or a conjunct cluster in an Indic script,
there are at least some users who expect Backspace to erase one element
of the cluster (the "last," whatever that means) and some who expect it
to erase the entire cluster. Each type of user will be frustrated if the
other behavior occurs.

> So that is the real conundrum here. Getting all of the commands
> of a text editor to work efficiently the way users expect is
> itself an art form -- even for relatively simple scripts. So it
> really should not be too surprising that people have rather
> intense arguments about how such operations *should* work
> for abugidas. (Particularly because such operations very
> often are not occurring in monolingual/monoscriptal
> contexts, and expectations carry over from one language/script
> to another.)

Daniel Bünzli pointed out that French-speaking users would consider
'é' a unitary letter, and would expect Backspace to erase the whole
thing, even if "under the hood" it might be encoded in NFD as <0065
0301>. It's not at all clear that a Sinhala user would expect Backspace
to delete a cluster of three or four "letters" (Naena certainly didn't
like that). But the two scenarios are quite similar as far as software
and encoding are concerned. So maybe a French keyboard could have one
Backspace behavior built in, and a Sinhala keyboard could have something
different, something that may or may not be possible under various input
architectures.

--
Doug Ewell | Thornton, CO, USA
http://ewellic.org | @DougEwell