On 7/10/2014 5:08 AM, bearophile wrote:
(Sorry for the very late answer.)
Walter Bright:
Stack overflows are not safety problems when a guard page is used past the end
of the stack.
It's not a safety problem in Erlang/Rust, because those languages are
designed to manage such failures in a good way.
Please explain.
The idea comes from Erlang language (and perhaps Erlang has coped it from
something else), and then Rust copied it (and indeed, if you look at the
"Influenced by" list here, you see Erlang, and it Rust has copied only the
Erlang feature I am discussing here: http://en.wikipedia.org/wiki/Rust_language
).
Erlang systems must be extremely reliable, they run telecommunication systems
that must just always work, with only seconds or minutes of downtime every year.
But programs contains errors and bugs, so to face this problem Erlang (and now
Rust) has chosen a curious strategy.
The description, see "4.3 Error handling philosophy" at page 104-109, a PDF
file:
http://www.erlang.org/download/armstrong_thesis_2003.pdf
It's also a bit explained here, at the "3. What is fault-tolerance" section:
http://stackoverflow.com/questions/3172542/are-erlang-otp-messages-reliable-can-messages-be-duplicated/3176864#3176864
Some more technical info:
http://www.erlang.org/doc/design_principles/sup_princ.html
Bye,
bearophile
Thanks for the links!
