On 10/08/17 09:09 AM, roger peppe wrote:
I quite agree: I suspect it a tradeoff, in which the more you can easily
represent, the more you need to carefully architect exactly *what* you
On 10 August 2017 at 13:39, David Collier-Brown <davecb...@gmail.com> wrote:
On 10/08/17 02:47 AM, Henrik Johansson wrote:
I beg to differ. Most Java apps I have seen over many years almost
unanimously suffer from over-modeling.
A former customer did a deep, thoughtful, *thorough* model of bracket
tournaments, without any attempt to abstract the salient features. Java
represented it beautifully, in complete detail...
So it's now impossible for a single person to keep it in their brain, and
every attempt to change it introduces new, surprising "features".
In effect, the ability to represent anything easily led to our error: we
I remember something similar happening with some Haskell I wrote.
Because the type system was so powerful, it felt wrong not use it to
represent everything, which ended up problematic. Frivolous thought:
I wonder if there's an (far-fetched) analogy to be made between this
and dropout techniques in neural networks - if our type system is really
powerful, it's easy for our type structures to "overfit" to the current
problem being solved, making it less adaptable and maintainable when
the problem changes.
David Collier-Brown, | Always do right. This will gratify
System Programmer and Author | some people and astonish the rest
dav...@spamcop.net | -- Mark Twain
You received this message because you are subscribed to the Google Groups
To unsubscribe from this group and stop receiving emails from it, send an email
For more options, visit https://groups.google.com/d/optout.