[PRL] Benjamin Pierce: Types Considered Harmful

[Felix S Klock II]

On May 30, 2008, at 10:15 AM, Dave Herman wrote:

> The weirdest part was when he claimed that one of the big issues  
> people
> have with type systems is performance. Does anyone actually say this?

Most type systems (of which I am aware) require a memory safe language  
for their soundness proof to go through.  Many languages rely on GC to  
get memory safety.  People have argued against (tracing) Garbage  
Collection based on performance.

So I think that is the heart of the argument there, which can be  
summarized as: "if you reject GC due to its performance implications,  
then you might as well throw out the goal of type safety as well."

I do not know how many people actually make this argument to conclude  
that you should reject types; it requires a pretty deep knowledge of  
type systems and a somewhat shallow knowledge of GC.

The argument does require that you accept *every* implication on the  
title of slide 19.

(I reject a majority of the implications on the title of slide 19.)

-Felix