Re: RFC 87 (v1) Timers and Timeouts

>>>>> "UG" == Uri Guttman <[email protected]> writes:

CF> The need for timers is independent of AIO.

UG> no totally. i don't see any easy way to support multiple timers without
UG> loading in some parts of AIO. and yes, that is implementation but i see
UG> those things as i design the language level stuff and i can't hide from
UG> it. you would get multiple timers only if you get AIO as i am descibing it.

Sorry, I don't buy this.

Why can't I have multiple timers, and not be able to have AIO?

If you don't have both, then there is a design. If you have both,
then you would select a uniform api.

The non-uniform api would probably be simpler than the more generic
api.

If some sort of event loop is put into perl, I'd rather that there be
_no_ callbacks as part of the interface. I find the straight forward
loop easier[1]. And if necessary, an external module can implement the
callback style[2].

Having both just adds complexity.

<chaim>

[1] I've done both styles and written both style dispatchers also.


[2] I think the loop can emulate the callback style much easier
    than making the callback emulate an event loop.
-- 
Chaim Frenkel					     Nonlinear Knowledge, Inc.
[email protected]				               +1-718-236-0183