And another iteration:

Is it me, or is the mail-archive not showing all mails in a timely manner? It
took two days for one of mine to show up.

Anyway, to answer a few questions:

> Changing the selected wallpaper group is something the user will do a few
> times and there are a fixed, small number of choices.  Yes it would be very 
> nice to show the user that many of the groups are similar and to make the 
> user aware of the differences.  However, I think there may be other ways to 
> do this and they may even be more clear.

Yes, that's why there's a drop-down there. If you want to choose a tiling type
immediately, then you use the drop-down. Also, the tile type can only be
changed in tile edit mode. Users probably won't spend a lot of time in that

> In contrast, changing the size, orientation, in some cases angle of the base 
> tile with respect to the group being tiled is something the user will need 
> fine control over.

The tiles behave like objects, so you can transform them with the F1 transform
tool, and you can snap them to other guides. 

For radial and line tiling guides, you can also input numerical values 
in tile edit mode. I'm not sure what to do with the wallpaper tile though, since
there are so many different shapes. A possibility is to define the coordinate
of the first center of rotation + the length of 2 sides (only one is used if
all sides are equal length).

Now to answer Jasper:

> 1. Inkscape's current tiling approach doesn't actually generate the
> tilings it says it does. It /can/, but by default it doesn't (you need
> to tweak the parameters to make sure the tiling actually matches the
> chosen tiling group). 

Ah yes, I noticed the hard way that P2 tilings behaved weirdly... (i.e. I
was trying to understand what it does by actually trying to render it)

> Forget about tiles. Seriously. The problem with focusing on the
> actual base tile are two-fold. First of all, if you define the base tile
> together with what rotation centers/reflection axes are on its edges and
> at what positions, then it is often non-obvious what the actual base
> tile is. 

But the base tile is still the most intuitive guidelines one has for wallpaper
tilings... Also, you need to know where the edges are to create seamless 

Or do you mean that the extra stuff gets in the way? The handles
etc. aren't visible outside tile edit mode. Since the default tile is a simple
P1 tile, and you need to bother double-clicking the tile to enter the
tile edit mode and change what it does, I think we can assume that the
user won't be too surprised if the output is no longer a simple P1.

> Even worse, take a p2 tiling and put a 180-degree rotation
> center on its "translational" edges (I'm referring to the diagram drawn
> by veronika(?)), what you end up with is another p2 tiling with a
> different base tile... 

Huh? It's still the same tile, just flipped. That's how wallpaper tilings work.

> Another reason is that users might want to put guide points somewhere 
> else than on a tile boundary. So I think it is more important to focus on
> the transformations involved, rather than on the base tile, and I think
> this can even help to make things make MORE sense.

I'm not sure what interface you're proposing. I know nothing about tilings,
and the only thing I care about for tiling is that when I click the "apply"
button, I get a pretty result (and a seamless output would qualify for the
"pretty" factor). The specific transformations don't actually matter at all
to users like me. :S