Re: [Gzz] Summing up...

[Benja Fallenstein]

Tuomas Lukka wrote:

> > How about making the connections work again? The problem is not showing 
> > them in gl (I have a fix) but the coordinate systems: currently they're 
> > drawn between the ul corners of the cells, because of the (w=2, h=2) 
> > coord systems. Fix needs to be implemented in both VobVanishing and RowCol.
> >    
>
> So we want to move to the convention that Vobs are in (0,0)..(1,1)?
> Any opposition?

Fine, but doesn't change much here of course: the connections are still 
between ul corners.

The problem is that the connections assume that (1,1) transforms to the 
lower right corner of the rectangle, but with (w=2, h=2)-- or 
(w=1,h=1)--, it's still the ul corner, just one or two pixels away.

> > > Hmmmmmm.
> > >
> > > What I'm thinking is whether celltexter is the right interface; this goes
> > > again the same route as enfiladematching: is this robust and clear enough
> > > to be in the core.
> > >
> > > Because it might be better to handle this by the views, possibly through
> > > an intermediate interface.
> > >
> > >      
> > Hm. That means that the insertText() etc. functions won't work with it. 
> > Isn't the point of celltexter to hide the implementation details of 
> > text, presenting "the text in a cell" in a coherent way and allowing 
> > editing of it through a set of methods that take care that the internals 
> > (xu identity etc.) aren't broken?
> >    
>
> Yes, but here there are more implementation details than usually.
>
> Hmmm.
>
> How about keeping it outside the core for now, making an implementation
> on top of it, and then, later, we can move it in once it's really
> solid.

Ok.

> I really want us to understand all the tradeoffs. Also, how much 
> policy would the core need to handle for this: e.g. on text 
> insertion, how does it decide which cells things go into?

Yes, true, I don't know yet either.

What do you think about my proposed mechanism to do paragraphs?

- Benja