[erlang-questions] binary_to_term and leaking atoms
Tue Jan 5 01:11:23 CET 2010
----- Original Message ----
> On the down side, atom
> GC sounds like it could easily kill performance. I'd imagine that everything
> would have to stop when atoms were GC'd.
In my opinion, full stops are probably not needed (well, hardly ever). I furthermore think the overhead can be low, e.g., normally not noticeable compared to today, though there are currently no real numbers to back that belief up. (see http://portal.acm.org/citation.cfm?id=1088369 for details)
> Even if not, I'd also imagine it would
> be a large and invasive change.
The atom GC described above would as far as I can see mainly affect scheduling, code loading and regular garbage collection. But regardless of invasiveness, atom GC plugs a hole that leads to trouble (as we can see on this list). I think it's worthwhile.
More information about the erlang-questions