Re: commit: Cut and paste MathML

On Wed, 2005-02-02 at 11:32 +1100, Martin Sevior wrote:

> > Martin, please don't do that. We already discussed the fact that we
> > should stop using RTF for cut&paste internally anymore.
> >
>
> Sorry Hub. I explicitly stated that I NOT agree to that.
>
> What I said was, your time is your own. There are lots of difficult
> issues involved in migrating cuts and paste out of RTF with significant
> costs.

OK. Let's get started on that.

-RTF conversion can cost too much in time. Yes, not everybody use a
multi-GHz CPU on his computer. My fastest machine is still my 400MHz G3
PowerBook. And I have seen case where RTF conversion was somewhat slow.

-Having a bi-directionnal RTF conversion without loss has proven to be
hard, even more of the data model we use is different than of RTF. And
this is what we does, and this won't get better.

I have been advocating that change, to NOT use RTF as a cut&paste medium
INSIDE AbiWord, for a long time, for that exact reason.
To achieve that, we should implement some object persistence for PT
objects, and be able to insert them whenever needed. Shouldn't be hard,
and would be a good occasion to sanitize a little bit.

>
> I do not intend to spend my development time doing it.

But you do spend time preventing people from doing it. I still haven't
dug into PieceTable because each I have head-ache understanding
*uncommented code* that seams to be interacting with everything else.

> > What you are doing is just making our life harder to get read of that.
> >
>
> I do not see how this makes your life any harder and saves me at least
> one month of development time.

You see *short term*. I see *long term*. As said above, this feature
would take some time to implement, but once done, it would save A LOT of
time because then RTF would only have to stick to the spec and eventual
third-party crap, but not our own.

"abilist" is a good example of what I dislike with that cut&paste.

Hub

-- 
Crazy French - http://www.figuiere.net/hub/