David A. Ham

We've had an offline review from @wence- and we think that, while the interface is right, we are wrong to think of this as a terminal. In fact it should...

Hi Jack, It's "external" in the sense that it's external to UFL. It's a foreign function interface. I think that that's a better description than the other suggestions you make,...

I think this looks fine. I have started https://github.com/firedrakeproject/firedrake/pull/2568 running just to make sure.

> I think this looks fine. I have started [firedrakeproject/firedrake#2568](https://github.com/firedrakeproject/firedrake/pull/2568) running just to make sure. So it turns out that the sobolev spaces change (which hadn't yet been merged to...

Sorry for taking a while to get back, I'm catching up after being on strike. The reason why dual coefficients and dual arguments need to be different types to primal...

Hi Rob, This is actually a disgusting design flaw in UFL. It has its own type system layered on top of the Python one. I think this was done because...

> @ksagiyam Can you please explain again what is this about, but in terms of a pure mathematical language? I am still having difficulties to understand the mathematical concept behind....

I think that the original question here misses the point. @mscroggs is trying to get Dolfin to recover lost information. Recovering lost information almost always indicates that the design is...

I'm in favour of backward-incompatible API improvements in UFL. However I'd rather they came in a limited number of batches, rather than a constant dribble. Users should understand us changing...

Isn't the current behaviour simply the expected behaviour because `*` is left-associative? Why is that an issue?