> SWM> Question: Can the interpreter determine when a variable becomes
> SWM> shared?
> SWM> Answer: No. Then neglecting to put a :shared attribute on a shared
> SWM> variable will crash the interpreter. This doesn't seem very Perlish.
> Err, no. It won't crash the interpreter. It'll make the script operate
> incorrectly.
This is just the distinction that I am concerned about.
In RFC 178, I used the term *data coherence* to mean that the
interpreter won't crash or corrupt its internal data representation,
and the term *data synchronization* to mean that the program actually
does what the user wants. Perhaps we could use the terms *internal*
thread-safety and *external* thread-safety, instead.
Here is an example of what can happen without internal thread-safety
Perl source:
Thread1 Thread2
$a = 'a'; $a = 'a' x 1_000_000;
Exeuction trace (C language pseudo code)
a.size = 1;
a.data = malloc(a.size);
a.size = 1000000;
memset(a.data, 'a', a.size);
Crash.
Crash & Burn.
Do not pass Go,
Do not collect $200.
Here is an example of what can happen without external thread-safety
Perl source:
Thread1 Thread2
$a = 'a'; $a = 'a' x 1_000_000;
print "Thread1 $a";
Execution trace (Perl code)
$a = 'a';
$a = 'a' x 1_000_000;
print $a;
and the user gets 999,999 more characters of output than they expect.
If the users cares about external thread-safety, they have to do their
own synchronization
Perl source:
Thread1 Thread2
lock $a;
$a = 'a'; $a = 'a' x 1_000_000;
print "Thread1 $a";
Now the output is guaranteed to be `a'.
All I want the language to guarantee is internal thread-safety.
Everything else is the user's problem.
- SWM
