On Thu, Jan 17, 2013 at 8:30 AM, Eli Barzilay <e...@barzilay.org> wrote:
> A sandbox can be relatively expensive, but it should definitely be > much cheaper than a new process. > In terms of time, I guess so, but in terms of memory, since the old process dies, the only real cost is restarting time (which should be rare enough in my case). > Also, bear in mind that the code snippet that Tony showed is very > partial: Racket has lots of OS-like features that need to be done to > get a more complete separation of the sandboxed environment, and his > code only some of that. The sandbox library is, roughly speaking, a > place to put all of the required features together. And if you're > worried about size -- one thing that you can do with namespaces is > share module instantiations so that you only get "billed" for them > once, and you can do that for both modules that you require, and for > your own modules if there is no harm in sharing state between > different sandboxes. (Setting up such sharing can be a little tricky, > but with the sandbox library you can easily do that with the > `sandbox-namespace-specs' parameter.) > Thanks for the information, I think I'll stick to the reload process for now, but I keep that in a corner of my mind. Laurent
____________________ Racket Users list: http://lists.racket-lang.org/users
