I've really wanted @threadsafe for a while for methods that are safe so that you can have compiler errors when calling non-threadsafe APIs. Kevin
On Fri, Jan 9, 2009 at 8:08 AM, Benjamin Reed <br...@yahoo-inc.com> wrote: > yes, that is a good article. it is actually the one we used to decide about > the current way of handling InterruptedException. in retrospect it turns out > to be a nice way to document that a call is blocking. > > ben > > -----Original Message----- > From: thomas.john...@sun.com [mailto:thomas.john...@sun.com] > Sent: Friday, January 09, 2009 7:56 AM > To: zookeeper-user@hadoop.apache.org > Subject: Re: InterruptedException > > Kevin Burton wrote: > > On Thu, Jan 8, 2009 at 3:21 PM, Benjamin Reed <br...@yahoo-inc.com> > wrote: > > > > > >> InterruptedException is rather tricky because the semantics of > >> Thread.isInterrupted() is rather vague. specifically, it is unclear why > >> someone would interrupt a thread. usually Thread.interrupt() is used to > shut > >> things down which requires special handing. thus we propagate it. for > >> example, i'm not clear how you shutdown your poll() method. an easy way > to > >> do it would be to use Thread.interrupt(). > >> > >> > > > > so if you just don't have it throw InterruptedException then > > thread.interrupt can't be used. It's an API decision really... if you > don't > > want people to interrupt then we don't have to throw > InterruptedException. > > I don't know many applications that use this in practice... does anyone > on > > this list? I always felt interrupt() was vestigial... notify/wait are > > somewhat in the same category IMO but at least they are useful. > > > > > See http://www-128.ibm.com/developerworks/java/library/j-jtp05236.html > for the idiom associated with thread interrupt methods and > InterruptedException. I will hasten to add that the appropriate use of > this idiom in various libraries (including in JDK libraries) is > inconsistent at best and the behavior is in some cases OS dependent. So > being careful about the use (maybe even stearing clear) of interrupt for > 'cancelable' operations is probably wise. > > In my poll method I don't shutdown.... the goal was to have the developer > do > > everything in an event API and code that way. > > > > Doing things in both sync and async operations is confusing. > > > > Kevin > > > > > > -- Founder/CEO Spinn3r.com Location: San Francisco, CA AIM/YIM: sfburtonator Skype: burtonator Work: http://spinn3r.com
