About GSM/Wifi handover

Wally Ritchie wally.ritchie at gmail.com
Thu Jul 26 23:10:04 CEST 2007

On 7/26/07, Alan Jones <junk at alan2.com> wrote:
> > It's called UMA and is documented in 3GPP TS 44.318 and 3GPP TS 43.318
> Well, with a quick look at 44.318, which is titled "Generic Access (GA) to the A/Gb interface", I'd say that this is not a spec which is going to help us out. 43.318 is the technical realization of the same technology.
> This diagram displays the location of the A/Gb interfaces pretty clearly:
> https://www.nethawk.fi/products/playground_1600x1200.jpg
> Anyways, the access to the MSC which those specs define is way lower (stack wise) than we have access to. If you look at figure 2 and 3 in 43.318, the stack is pretty clearly defined. The GA protocols just provide an alternate transport layer for the GSM Layer 3 messaging, on which messages from MM and friends travel, instead of going out over RR/LAPDM/GSM. This technology is more akin to a mobile handover between GSM and UMTS (Inter-RAT Handover), where all the signaling and mobility stuff stays the same, just the raw data transport (L1 and L2) changes.
> I can say with almost absolute certainty that we will not be able to get this messaging without explicit support from the GSM baseband module... And in that case, I would fully expect that the baseband module would implement the GA and IPSec stuff itself (maybe in a case where it owns the WiFi controller) for security reasons.
> Wally, you seem to be pretty familiar with 07.07, have you seen any extensions to support this technology from the high level commands?
I haven't seen anything specific other the EAP and definitely this
cannot be done without
support (modest I think) in the chipset.

At the R interface, it may require little more than a new PDP type for
the GSM Layer 3.
This would most likely be carried on a separate multiplexor channel.
Once a suitable secure tunnel is opened for transport, the GSM stack
can perform the
handover as appropriate. I'd like to think a bit more about the QoS
I think the point is that the tunnel is secure and the TE functions as
a connectivity
mechanism. Everything beyond the connection through the public
internet is managed
entirely within the GSM stack. It is relatively safe to export this
functionality through
the TE because everything is encrypted by the GSM stack. QoS is probably needed
end to end to keep the packet delays reasonably for the RTP payloads.
Link speed
to the chipset probably needs to be increased to lower latency as well.

> I honestly think that it will take a VoIP-based solution (supported by the service providers), since it is really the only true open telephony technology, in order to get the signaling and data transport to a high enough level that we could implement this sort of thing without explicit support in the baseband. With that said, from a network perspective, some level of support in the baseband would make things significantly smoother, so I would anticipate that any provider doing such thing would just expect the baseband chipset to implement at least some stuff. Keep in mind that generally, Handover is fully controlled by the network, with input from the MS.
I think that in relatively short order (in GSM terms) we will see
interworking with GSM
directly through SIP gateways and tunneling of GSM streams through the
public internet.
But I leave the non-technical discussions to a different group. What
is relevant here is
that we don't preclude supporting these in the architecture of gsmd or
its successors.

Cheers, Wally
> -Alan

More information about the gsmd-devel mailing list