Amazing that they use Yate (which is unfortunately not that widely known).
Years ago I've examined different VoIP engines for a project:
While Asterisk is one of the most widely used VoIP engines the source code and the whole architecture (at least few years ago) was ugly as hell. Instead of designing the whole system for concurrency there were some pretty ugly workarounds. In addition, it was trivial to crash the system with a little bit of SIP fuzzing (and some of those crashes may have been exploitable).
Yate on the other hand - while not having as many features as Asterisk - was just beautiful. A really nice architecture and extremely good readable C++ source code. In fact, it was a real joy to study the source.
Unfortunately I didn't really had a chance to study Freeswitch, as it was pretty new at this time and therefore not an option. From what I've seen it seems better architected and more stable than Asterisk. The main difference to Yate is that Yate only has few external dependencies, while Freeswitch tries to utilize as many external libraries as possible (e.g., instead of implementing SIP itself it uses an external library).
If I'd had to choose a VoIP engine today (and if a simple SIP proxy wouldn't be sufficient), Yate would be pretty much at the top of the list.
And the team behind it is really cool - met them on the 2.0 launch party and they were extremely friendly. They're honest about the project priorities too. It's basically a "what feature do you want to sponsor" project. (and I mean it in a good way)
It appears from the article that this is one half of the connection of Google Voice to SIP voip, the inbound sip call to gv number.
The other half, delivering GV inbound calls to a SIP phone, is still missing in action. It's also now getting a lot more attention due to the impending shutdown of the Gizmo5 sip service.
As much as Google seems to want me to switch my gizmo number to GV, GV solves a fundamentally different problem than I need solved. They are working on getting one number to any end point, and I need to map one number to my desk (without using a cellphone or hard wired phone.).
I'd love to find a good provider, but they all seem to be geared either to monthly fee for all inclusive, which is way more than I need for this application or not providing both an inbound number and the sip termination. (like ipkall) I could hack a combination, but it's probably going to wind up being less robust.
I tried to use a few numbers from IPKall but they were all already registered with Google Voice. Sipgate[1] gives you a number free and was easy to setup with some instructions[2] from XDA.
God, these addresses. I hope that one of the features between now and the actual rollout is letting the users assign much more memorable identifiers like 'username@voice.google.com". Still, the day when Android phones have out-of-the-box solutions for IP calling is getting tantalizingly close!
Blink (http://icanblink.com) is ok as a client. I'm not 100% on the interface, but it does seem to work well without crashing. It's gpl, and it does seem to have some momentum behind it.
There's one called simply "Telephone" which is very simple, just a textbox to call. No contact list or anything. (well, it wasn't much in december, since then they've moved from code.google to github: https://github.com/eofster/Telephone)
Gizmo5 is a crashy piece of ...something. It's the only app I've ever had on osx that has crashed so hard it's needed to be reinstalled. Several times. OTOH, you can't get it anymore, and Google is shutting down the sip service behind it.
You might try VoiceMac from http://mrgeckosmedia.com/applications/info/VoiceMac . I've not used their SIP support extensively, and it's primarily aimed to be a Google Voice client, but it might suit your needs.
I also don't know if the version on the page is the latest beta, or simply stable.
The developer has been quite responsive to bug reports, etc.
Google had enabled for a while a XMPP in/out gateway but they seem to have disabled it. It's still working on my account though and pbxes.org has a nice support for it built in, so it seems imminent they'll re-add this feature at some point in the near future. Given the on-deck support for SIP calling in Gingerbread, I figure they must be in the final stages of testing now.
If you want a relatively easy way to use your GV number for inbound and outbound calls with the SIP client of your choice, just connect it with Asterisk. The easiest way is to use "PBX In A Flash". I've been using it at home for a few months and it's been fantastic.
As for Yate, we believe it has the same issue we found with native Android SIP support: It doesn't distinguish between auth username and username, which creates limitations. (SIP Authentication username not configurable - RFC 3261 Section 22.1)
I've recently switched from SipDroid to CSipSimple (https://market.android.com/details?id=com.csipsimple&fea...) - which in my opinion works much better. The main difference is that SipDroid implements almost everything in Java code, while CSipSimple is a wrapper around PjSip (which is a pretty mature SIP library).
So if this doesn't work in SipDroid it may be possible that it works with CSipSimple.
Likely because it's significantly more expensive to call a UK mobile than a US one. I believe that they announced free calls until the end of the year. Maybe they'll roll it out in the UK at that point but expect to pay for it.
Twilio: $0.02/min to all US and $0.03 to UK landlines or $0.32 to UK mobiles -- yes, 16x the US rate.
Voxeo: $0.02/min to all US and UK landlines or between $0.14 and $0.35 for mobiles.
Google wouldn't use Twilio or Voxeo for this kind of service so it gets cheaper elsewhere -- especially on their scale -- but it shows the relative difference.
UK rates really are a crapshoot. Voxeo and Twilio must just be averaging the rates for mobiles. The rates I get at Flowroute for UK mobiles vary wildly: https://www.flowroute.com/services/rates/U/
... and the recipient also pays for "received airtime". Whereas, in most of the world outside the US, mobile phones get a different area code, and it is the caller that pays for "airtime".
To be fair, its not that bad. When I lived in the states unlimited nights and weekends started at 7PM and lasted 8AM. I also had Sprint SERO which was completely unlimited at $99/mo. + $25 in taxes.
When I moved to France, unlimited and unrestricted calling plans didn't exist yet and the talking time was pretty measly. Even now their nights and weekend billings were awkward 8PM-Midnight, Saturday afternoons - Sunday Midnight. They gave you a feature that was practically worthless.
This explains why SMS took so long to get off the ground there. A message cost $0.15, which is about the cost of a voice call, and unlimited messaging plans were $15-20 in addition to your voice plan. That money could be better spent on an even bigger plan with 1000 minutes. That and T9 sucked as a keyboard interface.
Eventually, I hope the rest of the world gets real
There are a handful of internal users w/ UK numbers but cost has been the concern since it's significantly more expensive to acquire numbers in the UK vs. the US.
Years ago I've examined different VoIP engines for a project:
While Asterisk is one of the most widely used VoIP engines the source code and the whole architecture (at least few years ago) was ugly as hell. Instead of designing the whole system for concurrency there were some pretty ugly workarounds. In addition, it was trivial to crash the system with a little bit of SIP fuzzing (and some of those crashes may have been exploitable).
Yate on the other hand - while not having as many features as Asterisk - was just beautiful. A really nice architecture and extremely good readable C++ source code. In fact, it was a real joy to study the source.
Unfortunately I didn't really had a chance to study Freeswitch, as it was pretty new at this time and therefore not an option. From what I've seen it seems better architected and more stable than Asterisk. The main difference to Yate is that Yate only has few external dependencies, while Freeswitch tries to utilize as many external libraries as possible (e.g., instead of implementing SIP itself it uses an external library).
If I'd had to choose a VoIP engine today (and if a simple SIP proxy wouldn't be sufficient), Yate would be pretty much at the top of the list.