Re: Re: Look ma, no hub?

["Paul Robinson" <ukcueman@xxxxxxxxxxx>]

Strikes me the real problem is the requirement for *all* apps to comply.
You need to add
this new knowledge to all apps you're going to run; existing linux apps
would all need to
be updated.

If you wanted to run just one app that didn't implement this right, it
would fail. And to
work around that, you'd have to kill all your currently-running apps and
them make them
run with a hub.

I don't think the benefit-cost ratio is particularly good here.

Paul


----- Original Message -----
From: "Mark Hindess" <xpl@xxxxxxx>
To: <ukha_xpl@xxxxxxx>
Sent: Thursday, November 17, 2005 7:54 PM
Subject: Re: [ukha_xpl] Re: Look ma, no hub?


>
> On 17 November 2005 at 19:44, Jean-Paul Figer <jfiger@xxxxxxx>
wrote:
> >
> > From an architecture point of view, the hub is a key component of
a
> > distributed Home automation system. Being able to interconnect
X10, Asterisk
> > PBX, Winamp, my old home automation system, text to speech, IR
and more in a
> > fast and coherent way is a strong advantage. The hub concept is
more
> > important than the Xpl protocol.
>
> Please explain what you can do with a hub that you couldn't do
perfectly
> well without one if your system (like Linux) supports binding to the
> same port by more than one application?  If they can do this, then all
> applications will receive all the messages that would otherwise have
> had to be forwarded by the hub.
>
> I think that from my testing that anything you can do on Linux with a
> hub will work just as well without one.  I'd love for you, or anyone,
to
> point out precisely what functionality I'd be missing if I didn't run
a
> hub.
>
> > Please keep the hub.
>
> Why exactly?  Personally, I'm usually quite glad when I can remove
> something that is a single-point-of-failure.  I'd have to have very
good
> reasons to justify keeping it.
>
> Gerry has some good points about platform independence.  However, I'm
> still not convinced since it is trivial for any platform independent
> application to test if it is possible to bind to the same port twice.
> The application just has to:
>
>   1) create a socket with the reuse addr socket option
>
>   2) bind to port zero - so the system allocates the socket an unused
>      port
>
>   3) create another socket with the reuse addr socket option
>
>   4) bind to the port that the system allocated in step 2
>
> If any step fails then reuse addr is not supported and it should use a
> hub.
>
> A similar test could be added to any application - for instance, a hub
> could do this and refuse to start if it is not required.
>
> Personally I'd not have a problem requiring applications (and hubs) to
> perform this test on all platforms.
>
> Sorry to cause trouble here, but I think it is important to justify
> having single points of failure.
>
> Regards,
>  Mark.




___________________________________________________________
Yahoo! Model Search 2005 - Find the next catwalk superstars - http://uk.news.yahoo.com/hot/model-search/