Garbage collection

Memory has to be given back, or the program runs out. Manual management worked until programs got big enough that a single forgotten free was a production incident — so languages started taking the job away instead: LISP in 1958 traced the roots, found what was reachable, and freed the rest.

The modern argument is about pauses rather than leaks. A tracing collector can find garbage in microseconds with a write barrier, or in minutes with a nursery, or not at all with a concurrent server; the language’s job is to make the choice invisible to the programmer — which is exactly why Rust, the loudest exception, is such a useful counterexample.

The trade-off

• Collection trades pauses for leaks: a collector can make your memory usage unpredictable, which matters when you have 64MB and a latency budget.
• Manual management trades leaks for pauses, and puts the burden on every programmer — most of whom are not thinking about it.

Where it came from

LISP, 1958 — garbage collection, recursion and conditionals arrived together, because McCarthy’s machine had no room for a programmer to manage memory by hand.