Re: Considering a sig() refcount rollback

I second that.


--- Scott Wiersdorf <scott@xxxxxxxxxxxx> wrote:

> On Mon, Aug 28, 2006 at 02:10:51AM -0400, Rocco Caputo
> wrote:
> > Does anyone prefer that signal handlers keep sessions
> alive?  This is  
> > your opportunity to explain why.
> 
> I'm a weak POE user, and don't have a good grasp of the
> internals, but
> I could conceive of some scenarios where you would want
> this kind of
> behavior. See below.
> 
> > I don't want sig()'s semantics flapping back and forth,
> so a
> > reversion will most likely be permanent.
> 
> And I agree with Rocco; the default semantics should be
> "fixed"
> (whatever that means) and then left alone. May I second a
> proposal I
> heard last week (but have forgotten who proposed it)?
> 
> The proposal was that a special flag to sig() (or a
> separate method
> entirely) be given which could alter the semantics so
> that sessions
> are kept alive as long as there are handlers "out there"
> expecting
> that session to catch the signal.
> 
> Perhaps Rocco can revert the behavior as it used to be,
> and then in
> some future date he (or some other ambitious person)
> could add in the
> optional "keep-alive" semantics if it were popular enough
> to warrant
> it.
> 
> Scott
> -- 
> Scott Wiersdorf
> <scott@xxxxxxxxxxxx>
> 


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com