News for hackers. Stuff that matters.
Re:Fearless concurrency with Rust by steveklabnik on Friday April 10, 2015 @02:14PM (#9356093)
> It's so simple,

I think that in hindsight, this may be true, but it took a _very_ long time to figure it out. It took _years_ of iteration before we got to where we are today, and the previous work in this space, which used regions but not borrowing, didn't go further because of the ergonomic issues.

In retrospect, hard things often seem obvious, but that's because the work was already done!

Re:Fearless concurrency with Rust by MichaelGG on Friday April 10, 2015 @03:38PM (#9356640)
Exactly! I came to Rust thinking I knew all I needed due to trying to solve the same things by hand in the CLR. Like I figured I had a pretty good idea what I needed to allow F# to have fast memory management.

And while the core idea is correct, holy shit there's so many details that the Rust team has had to deal with and iterate over and reduce until it's nice. I suppose this is why the CLR and everyone just punt on the whole issue.

Re:Fearless concurrency with Rust by mkehrt on Friday April 10, 2015 @03:37PM (#9356637)
I'm pretty sure Cyclone used both regions and borrowing in ~2005. But it was a research project and not meant for wide use.
Re:Fearless concurrency with Rust by steveklabnik on Friday April 10, 2015 @03:44PM (#9356683)
I thought Cyclone only used regions, but I haven't read the paper in a while. (and dates from 2002 IIRC)

It certainly had a big influence on Rust, we're fans of the work.

Re:Fearless concurrency with Rust by cwzwarich on Friday April 10, 2015 @09:51PM (#9358079)
Cyclone definitely had borrowing. See section 3.1 of this paper: https://www.cs.umd.edu/~mwh/papers/ismm.pdf. Generally speaking, every attempt to use substructural types for resource management in a practical language has allowed for temporarily treating a restricted resource as unrestricted.
Re:Fearless concurrency with Rust by mkehrt on Friday April 10, 2015 @04:11PM (#9356838)
It looks like you're right.

I had an undergrad senior thesis in 2006 that futzed around with regions and borrowing, and I know I read some papers on their interactions, but I'm not immediately finding anything chasing references.

edit: Of course, I don't mean to diminish the accomplishments of Rust, which had turned these moving parts into a working language. In particular, Rust makes it easy to ignore these issues most of the time.

Re:Fearless concurrency with Rust by steveklabnik on Friday April 10, 2015 @04:43PM (#9356995)
Absolutely no diminish-ment taken.

To put it in startup terms: other people working on your idea only validates it, it's not a threat. Execution is ultimately what matters. I prefer letting academia blaze the intellectual trail, and then have a number of people implement it after. :)