I'm a bit confused about the title now... Are the "New dynamics interface
proposal" discussions about a completely different tool, or about the 
dynamics for the tiling tool?

By the way, Veronika, I keep losing the link to your list of tilings picture
(because I delete nearly everything and copy entire new chunks of
description each time I update with a completely new version)
Could you perhaps upload it to the Inkscape wiki?

Anyway, I updated the wiki with the latest iteration for the tiling tool:

http://wiki.inkscape.org/wiki/index.php/Tiling_tool

I've also added the "implementation steps." At this point, can everybody
take a look and see if they agree with the "core" interface (meaning
base tile + render frame at least, dynamics are still wip)? If so, I can
already put up the page as a blueprint so anybody interested can
get started (a lot of steps are still needed before reaching on-canvas
dynamics editing...)

Should I add a table to note what parts of the feature are accepted
designs, what parts are still under discussion?

Back to dynamics:

http://www.tomlechner.com/randompics/finishing.png

I was confused for a while, but after a better look I do really like
the interface! The term "wubwub" made me laugh. There should
be a way to choose and see which wubwub does what, and a way
to set the value back to zero. Double-click to add a corresponding
wub-wub? (since users are unlikely to need jittering Everything
at the same time)

Of course the problem is still how all this will be programmed.
Since you made that awesome alignment tool, I've no doubt that
someone like you can pull off a tool like this, but I'm not sure the
average volunteer student for inkscape could program this
type of interface. And we also face the problem of how to integrate
this with the existing mechanism.

Now, I know nothing about programming. Has anyone looked at
the tiling dialogue code yet to see how hard it'd be to mess around
a bit with it? Possibilities to add extra stops etc. even? Replacing
the opacity/blur/color mechanism with a gradient-like approach?

Right now the two possibilities for on-canvas tile editing are still:
1. Drag a tile, all the other tiles get updated dynamically
2. Drag a tile, a ghost copy gets created, but the other tiles stay
in place (Tom's proposal, and probably easier to handle than 
an interface where All the tiles are jumping around, preview
value or not)

I'd kind of like it if we can decided once and for all whether to use
the first or second approach though. So, any opinions?