Norm1 and NormInf, the great debate

[IainNZ]

Perhaps unsurprisingly, someone tried to use these and they didn't work. I think to implement them could require disabling hot starts (due to the complexity of mapping between MPB indices and the original JuMP problem), and would be another complexity increase (but how much I'm not sure)

[joehuchette]

Why don't we implement this first as an extension and see how it works?

[mlubin]

I'm very much against adding ad-hoc transformations like this. We're going to end up with a huge increase in complexity and a DCP-ish syntax that isn't as useful or consistent as Convex.jl. Either we do full DCP or we don't.

[joehuchette]

(Let's do full DCP)

[IainNZ]

Taste the rainbow

[IainNZ]

I like it being in a little JuMP tools package though

[mlubin]

JuDCP?

[iamed2]

Is this a thing that went forward?

[mlubin]

Not at all, should probably close this as too speculative

[IainNZ]

I made a start at this in https://github.com/IainNZ/JuMPTools.jl

[mlubin]

No reason to leave this open