Inferring from context declarations

George Russell ger@tzi.de
Thu, 22 Feb 2001 18:18:11 +0100


Marcin 'Qrczak' Kowalczyk wrote:
[snip]
> It would require to keep bodies of all polymorphic functions in
> a form allowing instantiation in different modules. For example
> ghc already requires too much memory (even a hundred MB for large
> modules), and it leads to code bloat, so in the current state of
> technology it's not advisable to always compile out polymorphism.
[snip]
I'll bet Marcin a pint (if he wants) that in 10 years time you will
need 1GB of virtual memory just to run the latest version of Windows
out of the box.  How Microsoft will manage to use all that memory,
I don't know (will we have irritating pop-up windows showing
videos of wizards telling us how to save Explorer settings?), 
but then 10 years ago I would have found it difficult
to conceive of a desktop computer needing 64MB.

The current GHC separate compilation strategy probably is advisable
at the current time, given available constraints, but I don't think
we should make language decisions on that basis.