[Haskell-cafe] Well typed OS

Joachim Durchholz jo at durchholz.org
Sun Oct 21 20:19:01 UTC 2018

Am 21.10.18 um 19:43 schrieb Vanessa McHale:
> I've never understood why functional (and in particular 
> Haskell-influenced) approaches to hardware never took off. I suspect it 
> was political (Haskell is too academic, etc.), or perhaps the companies 
> using it are just quiet about it :)

The usual reason is that they simply haven't seen it work, and are not 
sure whether it will work. Also, retraining your engineers is VERY 
expensive: expect a few weeks of training, plus a few months of reduced 
effectivity; multiply with an engineer's cost per month, and you arrive 
at really impressive figures.
If a company is successful, it sees no reason to go into that risk.
If a company is at the brink of defaulting, it cannot afford to try.
It's usually the second-in-line companies that try this kind of stuff, 
and only if they have some risk appetite and need/want to catch up. In 
these companies, there's room for experiment - but the experiment needs 
to show success to be adopted, which is by no means guaranteed (you can 
easily fail with a better methodology if you don't do it right).

And maybe functional really isn't a good match for hardware.
I don't expect that to be reality, but I have too little experience with 
functional and almost zero with hardware, so I may be overlooking stuff 
- the design space for hardware is pretty different than the design 
space for software, pretty restrictive in some ways and unexpectedly 
flexible in others.
Which brings me to another possible reason: Maybe functional isn't 
solving any of the real problems in hardware design. That would then be 
the "meh, nice but useless" reaction.
Just speculating about possibilities here; as I said: too little 
hardware design experience here.


More information about the Haskell-Cafe mailing list