[Haskell-cafe] Cabal failures...

[Albert Y. C. Lai]

On 12-11-19 04:25 PM, Gregory Guthrie wrote:
> I am not exert in the area, but I wonder how /why/ this is different than other package managers, like apt in Linux, I have never had any problems with it, and I would think that their dependencies are of at least similar complexities.

I feel very strongly about the dissonance in comparing problems without 
comparing costs.

Debian has a horde of volunteers for just the menial and manual work of 
perpetually finding one coherent set of versions so end users don't have 
to. And in practice, There is never one coherent set of versions. There 
is only a not-too-incoherent set of versions, and the volunteers first 
have to decide on it, and then manually pick patches from other versions 
("backporting patches", "cherry-picking patches") to turn the 
not-too-incoherent set into a coherent set that does not exist in any 
pristine version.

On top of that, Mark Shuttleworth actually pays money for Ubuntu to 
start from Debian and further test the set, pick some more patches, 
unpick some other patches...

How many hours and/or dollars are you willing to pay for the menial, 
manual, perpetual chore of identifying coherent sets of versions so 
other people don't have to? And if a coherent set does not exist, how 
many are you willing to pay for backporting patches?

At least I paid my 3 hours to explain some cabal stuff at
http://www.vex.net/~trebla/haskell/sicp.xhtml

Even the Haskell Platform, one very small set, costs volunteer hours.