Re: [dispatch] [SPAM] Re: SIP and GSM/UMTS with OpenBTS

[Tim Panton new <[e-mail]>]

On 12 Feb 2014, at 16:01, Dale R. Worley <[e-mail]> wrote:

>> From: Tim Panton new <[e-mail]>
> 
>> Here is an interesting presentation by Kurtis that will give you
>> some feel for what these systems do:
>> 
>> https://www.usenix.org/conference/nsdi13/technical-sessions/presentation/heimurl
> 
> Though he's incorrect that the first cellular network was in 1991, it
> was in 1979.  1991 is the first GSM cellular network.
> 
>> It isn't directly related to the standardisation effort but is a
>> great background info.
> 
> It's interesting work, but I don't see its connection with SIP at all.

As I said it isn't related to the standardisation, but rather to give some background
to the usage.

> 
> My assumption is that you're asking, "How do we mate the GSM
> over-the-air protocol with SIP for backhaul?", that is, how do we
> disconnect the cellular system from IMS and all that.  But if your
> goal is to deploy base stations commercially, you're going to need the
> call accounting infrastructure of IMS.


That's assuming a currently popular billing model. If all your customers were (for example) on an
all-you-can-eat tariff then you might not need IMS. Likewise if you only had 1000 customers, the
standard IMS solution is overkill. If you don't need roaming, then you may not want the full IMS stack.

So to take Kurtis's example there might be significant value in a phone service that allowed users to:
1) call within network (that's a potential 20 mile radius even for a single cell)
2) make emergency calls
3) receive external calls
4) all bound to a specific sim
5) sms to the world

- That could be associated with a small monthly fee to cover the infrastructure costs. 

Sure they'd like the full LTE experience - but (a - wifi probably does that better (b - LTE is too expensive.


The point is these systems don't attempt to replace existing services - they are generally put into
places and scenarios where the IMS 3gpp model doesn't work (for one reason or another). This is evidenced
by the fact that these places _aren't_ served yet.

I'm hoping we'll standardise the SIP interface towards the ITSP, either in the form of a set of recommendations
and possibly a new header field or two.
I'd also like see a standard (SIP? SCTP? SNMP?) interface between cells to help with handoff etc.

The point of standardisation is to allow many vendors to interoperate and ITSPs to offer service,
but not to restrict the use-case so much that
it exactly replicates something we _know_ isn't practical for these environments.

Tim.




> 
> Dale