[hydra-dev] Demo for astricon

[John Todd]

On Aug 2, 2010, at 1:33 PM, Tim Panton wrote:

> I'm catching up on the conference calls.
>
> I'd like to add some sort of API demo in the list.
> So perhaps have some sort of REST API exposed that monitors/controls/ 
> drives
> one of the demos.

I don't know what the plans for the API are going to be by the time of  
the demo.  Kevin?

> I also like the 'cheap hardware' machine remark that John made,
> so it would be good to have a hydra component running on something
> small, like a pogoplug or android. It doesn't have to do much,
> but to show it is possible for lighter servers to contribute.

This might be an interesting one if we can find a system that supports  
Ice already and would just be a "compile-and-run" kind of situation.   
The problem with the smaller devices is that they take forever to  
learn how all the little dependencies and whatnot go together, or  
trying to cross-compile and export.  Fiddly business.  I suspect  
nobody at Digium will  have the time between now and the demo to delve  
into that, but if after some code is laid down you think you might  
want to try it and you can supply a "black box" that supports it,  
perhaps it might be an interesting addition, even if it's just to hold  
it up and wave it around during the demo and truthfully say "we got it  
running on this, too."

> I also thought about some sort of viral propagation concept,
> this might need to be licenseing based (as with DUNDi) where
> the expectation is all hydra instances can (by default) talk to each
> other over the net.

This is a really great idea!  However, should we restrict to E.164?   
There are some pretty hefty considerations to be made about what data  
would be exchanged between systems, and what, exactly, would be the  
protocol.  DUNDi is the obvious choice for number-based meshing.  What  
about non-E.164?  SIP URIs don't seem to need any mesh ideas - they  
have the DNS, and that works quite well.  XMPP is in the same boat.   
Freenum.org/ISN is in the same boat.  So really, I suppose I've  
answered my own question - only E.164 numbers would probably be  
considered (or private E.164 numbers or sub-numbers, but those  
wouldn't be meshed publicly.)

This leads us back to IAX2, actually.  It's unlikely that an IAX2  
channel will be done by the time of the demo, though.  It seems that  
IAX2 precludes any mesh ideas, since SIP, while wonderful, typically  
requires quite a bit more attention to get working properly.  If it's  
going to be a you-betcha kind of mesh (i.e.: everything works  
correctly by default except for a single bit toggle that is default=no  
but says "You should really, really, really set this to 'yes'!") then  
IAX2 would be the most likely to actually result in calls being  
delivered without additional thought.

JT


---
John Todd                       email:jtodd at digium.com
Digium, Inc. | Asterisk Open Source Community Director
445 Jan Davis Drive NW -  Huntsville AL 35806  -   USA
direct: +1-256-428-6083         http://www.digium.com/