I’ll miss it. (I’ve had a lot of fun with it – see earlier blog posts)

SFA (as it is sometimes known) is/was unique in that it allows a 3rd party developer to control a skype experience without being installed on the client’s desktop. It runs server side, in the rich environment of the Asterisk developer ecosystem.

When I wanted to bridge in chat from a different IM system, SFA was there, chat messages turned up as Asterisk events and I could manage them with my normal tool set. All I had to do was buy a license, no qualification, no pre-approval of the application, no certification of the hardware, just a creditcard, a click-through license and away you go.

SFA did presence properly too, your Skype status was available to asterisk, so calls could be routed to skype if you were online and to your cellphone if you weren’t (Skype can already do that you say – true enough – but with Asterisk I could whitelist callerid numbers and times of day, so that only the few would get to my cellphone out of office hours.)

So what went wrong?

In retrospect it seems clear that the wheels started falling off almost before the product got rolling. My guess is that Skype had a change of heart between the project’s announcement and beta. I don’t know if this was due to a change of direction or because they slowly realized what they had agreed to and wanted to minimize the damage.

The first hint that something was not quite right was when we applied to the Beta program, we got accepted, but I was told that there had been resistance from Skype.

The development of the product took ages, I really don’t know why, but the delay dampened enthusiasm.

Then, to my astonishment, Digium’s Mark Spencer – the instigator of the deal – started complaining publicly about the limitations skype were putting on the usage of the software. Only Skype IDs generated through a business account would be accepted for Skype for Asterisk, ruling out the easy transfer of existing normal IDs. We all complained but Skype were unwilling to budge.

When the product was finally launched, the license conditions included a provision explicitly banning the use of SFA in a ‘hosted’ environment. So you couldn’t offer SFA as a service, it wasn’t a matter of price, you just couldn’t do it.

So Skype deliberately hobbled SFA. What were they scared of ? Openness I think. If SFA had been released as first announced without the absurd limitations, Fring could have used it to run their skype gateway without getting shutdown. Phonefromhere.com could have offered a web-based skype client. Someone could have offered a REST interface to skype chat.

This might have made sense if Skype were planning to offer services that competed. But no, they simply didn’t want any software developers in their eco-system.

I’m not one of those who thinks that Microsoft ordered the axing of SFA, Skype probably never envisaged renewing, so when it came due, they pulled the plug.

However this is going to make for an interesting culture clash with Microsoft. For all their many faults, once they release a product Microsoft do encourage developers to use it to the full.

Anyway, now it is gone, leaving a bitter sweet aftertaste as a warning to future Skype partners.

For about a year now, we have been running a Skype gateway so that people who couldn’t or didn’t have time to set up a SIP device could call in (free) via skype.

Like most conference calls, it works best with a text back channel so that URLs, part numbers, email addresses etc don’t need to be laboriously spelled out. For historical reasons this is an IRC channel on freenode.net . So a typical user would have a softphone fired up on their laptop for the audio and an IRC client for the text.

Skype users are very loyal to skype chat, so a few weeks ago I started looking at trying to bring the IRC chat to them in a more convenient way. Last week I got an experimental system running that did just that. Here is a sample  of the result:

A skype chat bridged to gtalk via irc

In some ways that looks like an ordinary skype chat. Indeed for the skype user, that is exactly how it behaves, except that messages from the vuc.me user are in fact from multiple IRC users.

If you look a little more carefully, you’ll see that there is discussion of a gtalk connection (google talk). That’s because I’d also made the system support connections from gtalk too.

So users from all three of the ‘islands’ in the VoIP world (SIP, Skype and GoogleTalk) could all participate on an equal basis in a conference about VoIP. Not only that, but their IM messages were also exchanged freely, all with the correct attribution.

How was this done?

Asterisk is the key ‘glue’ here, with it’s support for Skype and Gtalk channels. But even Asterisk doesn’t support the kind of gatewaying we needed for the IRC . For that we needed the (excellent) asterisk-java package which communicates with Asterisk and allows you to manage and control calls in Java. I used the PircBot package to implement an IRC connection in Java.

As this was a prototype, I didn’t want to get bogged down in writing the gateway in Java, I wanted to use something more expressive and easier on the keyboard, so I used Groovy.

The end result was a very few lines of Groovy, here is the meat of it (sorry Jay) :

Now the fun and powerful thing about this code is that it is asynchronous and event driven. Each event from Asterisk is checked to see if it triggers any of our desired actions, but this triggering is done on a fine grain – in the last block we are only interested in Skype chat messages that are to the skype id ‘vuc.me’. We can write expressive filters in code to do the matching, then define the action to carry out in just a few lines. The over all effect is like the syntax of unix’s AWK pattern matching language.

This is quite different from the kind of scripting you see in twilio, asterisk dialplan, tropo and other telephony scripting environments. They are all about call control, this is about call enhancement. A traditional telephony scripting language sets up the call, but once the 2 (or more) parties are talking they step out of the way. Here I’m adorning a live call with extra asynchronous information.

There are systems that do this sort of adornment – ZIP DX sends messages to the SIP clients throughout a call allowing phones to display the name of the current speaker – but that’s more of an endpoint feature (see also the very cool thrutu). In this case the adornment is taking place in the middle of the network and in the middle of the call – and thats new (at least to me).

I’ll be polishing the prototype and hopefully have it deployed in a stable basis in time for next Friday’s VUC.

Then there was a long silence. I was a bit annoyed, I’d contributed my time and thoughts to the opensource project and the only place I saw any results was in their closed source Switchvox product.

In Germany at Amoocon (I blogged about it here too) , Digium’s engineers sat and listened stoney faced as I gave a talk which described how much easier it was to build a fault tolerant solution using the arch-rival FreeSwitch. But still they said nothing.

A few months later I found out why. Digium’s engineers had come to the conclusion that we had asked for the impossible. To meet our wishes, they would have had to re-write Asterisk from the ground up, while a million servers were still running it and all their users depending on it. Also once our wishes were granted it wasn’t clear that the resulting  product would be suitable for all of the things that Asterisk does today.

So they made a very brave decision, they’d have to start again. The new product (now known as Asterisk Scalable Communications Framework) would be a companion product, to work along side the Asterisk we know and love.

Earlier this year a group of some 30 community members were invited to Digium HQ in Huntsville for a mini-summit. I was one of the very few who knew what the summit was about, which made for awkward bar-talk the night before and put me on at least one journalist’s “Annoy list” (Hi Doug!) . There they presented the choice they had faced and the decisions they had come to. They asked us what we thought, they asked us what the priorities were, they asked for our help. They asked us to keep silent a little longer.

Astricon in October 2010, they held a keynote with an uninspiring title  (astricon keynote video) where new project was revealed. They invited 3 community members to be part of the annoucement and to help explain the background. (My bit is about 10 mins in) – They also released all the source code they had so far, along with the working wiki that afternoon.

As a result of all this careful consultation and community involvement, the announcement was met with a ‘that makes perfect sense, when can we start?’ attitude from the 700 developers and business partners in the hall.

Contrast this careful step-by-step involvement with the way Skype treats their ecosystem (like Fring and Nimbuzz ). Don’t get me wrong, I’m sure Skype had every legal right to block them, but I’m not convinced it was smart.  As a developer, I constantly find myself confronted with things I can’t do because Skype won’t let me or tell me how.

Now Skype have sprung the hated Mac Beta and the Facebook enabled Windows version on a shocked ecosystem. Did they consult with developers or partners ?- No I don’t think so.

The only counter example is the way that Skype have made the Silk codec available to the IETF as a draft standard and licensed it reasonably freely to developers.

Why should Skype care what I think? Let’s face it, telephony is long overdue a shakeup, the phone call may be ‘dead’ but voice communications over a distance isn’t. Indeed, it is still the killer app. Despite the total lack of innovation in telephony for 50 years people are paying (yeah, paying – hear that Web 2.0? )  billions for phone calls. So where is that innovation going to come from ? Will Skype’s hundred or so brilliant inward focussed engineers out-think a 70,000 community like asterisk.org or the wider startup world? I know where I would put my money.

Here’s a final thought, if you look 2 or 3 years into the future, you can see that realtime voice and video communications will be baked into the browser on every platform (see IETF discussions). Why would you need a fat (some say bloated) stand alone communications client then? Then a user’s choice of who’s minutes to use will be down to the quality of integration into your web experience.

“Look after your developers and they will look after you” as an escapee from Symbian put it.

Here are some recent examples I have encountered:

• Panasonic Skype enabled TV to Skype on a Mac – Video doesn’t work.
• Skype for Asterisk calls from Verizon and Three mobile phones won’t connect.
• Skype Connect can’t call out to Skype users (although the other way around works).
• Multiway Video calls only work on PCs.

Skype need to deal with this fragmentation before the cracks get too visible. Otherwise people who aren’t using Skype on a Windows PC will start to feel like second class citizens.

[ update: The Panasonic  to Mac issue was reported to me by James (too-many-devices) Body.

Jim Courtney corrected my description of the Verizon issue - I'd got it the wrong way around.]

If you look at it from a consumer’s point of view, this is a huge win. Imagine sitting in your lonely hotel room on a business trip, you call your family on your iphone. You check that now is a good time, then switch to video and loose all the costs (except perhaps the hotel’s wifi charge, but your company has paid that). This is pretty much how skype works now, except that there you require a latptop with skype installed instead. The flaw in the skype experience is the far end availability, your family may not be sitting in front of the laptop leading to a ‘not there’ problem (hence Skype’s interest in Skype for TV).

It is clearly in Apple’s interest too, since it generates a network effect around iPhone 4, owners will apply peer pressure on family and close friends and encourage them to buy iPhones so that they can get more FaceTime.

So where does this leave the carrier ? On the face of it Apple are doing them out of minutes for no return. Surprisingly the carriers are probably going to be pretty relaxed about it.

The loss of income argument only applies to Pay-as-you-go subscribers. Most iPhones are on monthly contracts, and very few people on contracts exceed their monthly minutes. So there is no incremental loss to the carrier of a shorter call, they get the monthly fee eitherway. Indeed, FaceTime unburdens the call from the carrier’s network onto the public internet (thanks to  the hotel wifi and the family’s ISP).

In effect, FaceTime acts as a portable, zero-install femtocell, offloading call traffic for free, without any carrier buildout or support costs or any customer setup and hassle. See Dean Bubley and Andy Abramson ‘s discussion on the merits of femtocells vs WiFi.

Ah, but what about those high-value roaming minutes the carrier is losing from foreign trips? Again the carrier has reason to be cheerful, the initial connection is made over GSM, so they get a connection fee (30p minimum?) and a minute or two of call at roaming rates. Consider the alternative, the smart business traveler these days uses Skype or Truphone or Fring, all of which earn the carrier nothing at all and draw them away from the carriers by using a new identifier – either a skype id or a new virtual number.

Which brings me to the remaining reason that carriers will quite like FaceTime, the id it uses is your current mobile number, not your apple-id – keeping you loyal to the numerical side of the great namespace divide.

All in all not a bad deal for a hard pressed carrier like ATT or O2, at least in the short-term.

For those of you who don’t know,  Ada Countess of Lovelace was (arguably) the first computer programmer, she designed an algorithm for Charles Babbage’s Difference Engine in 1843 (according to wikipedia).

She was an extraordinary woman, exceptionally rich, well educated and clever. She was often ill as a child. Her father (Lord Byron) was hardly conventional, these factors combined to give her the opportunity to explore and develop her mathematical abilities, a thing that very few women of her time (however rich) were allowed to do.

That’s the thing I want to celebrate, women in computing who have been allowed to develop their skills. Everyone can’t be the first, or have a turgid programming language named after her. Traditionally these blogs talk about Grace Hopper, or the women of Bletchley Park. They were exceptional, I want to celebrate the ordinary, women who turn up to work with computers and do a good (or excellent) job, and are judged on their achievements, not on the fact that they are women.

The first time I saw a computer in real life was some time in the early 70s when my aunt took me into her office in the Department of Education in London and showed me a modem and the terminal that connected it to the (gasp) mainframe. The department used the computer to try and predict the future demand for school places for planning purposes.

(That isn’t my Aunt Doris by the way, and her mainframe was probably an ICL not an IBM)

Throughout my computing life there have always been women around, using computers as a tool. I first came across the issue of women in computing when my university lab partner (Kate) and I signed up to do a computing project. The supervisors spent the first 2 sessions discussing how we would ensure the results didn’t put off women. We just wanted to get on with the project get the marks and get on with life/beer drinking. So I’m afraid we switched supervisors to a guy who set us a straight forward image processing problem and let us get coding.

Later on I worked with Gabriel  on a simulation of a chemical plant  - she is one of the brightest mathematicians I have ever met. Recently I’ve worked with Ilana, who has a computing track record going back to the 70s when she was an ‘analyst’ for Control Data and worked with Seymore Cray , but more recently HP, Microsoft, Cisco have used her skills.

While in Holland over 10 years ago I started working with Birgit, who I still share an office with, we have worked on loads of hard projects together, she’s got her name in an RFC and implemented a codec from the ISO standard description, as well as delivering consistent clean code day in day out.

That’s where it all falls apart, my gender blind dream world. VoIP. The VoIP world has inexplicably few women coders in it, so much so that I can only think of one who I regularly meet at conferences (and enjoy a fierce debate with). I’m sure there are a few I don’t know and Diana almost makes up for the lack with her determination and achievements.  But even trawling my email and twitter lists, I can still think of only a very few.

However much we might like to think the western world is gender neutral, it just isn’t yet. Last night my daughter asked me to buy a new ‘First person shooter’ game. I asked her if she could borrow  it to see if she liked it , she said no – none of the girls in her class had it and she didn’t want to ask the boys who did, I’m still wondering what to do.

It seems to me that there are two ways to level this playing field (to reuse a cliche), one is by making the argument – publicising the extra-ordinary. The other is a proof-by-example, where the extra-ordinary becomes the normal. Which you choose depends on how extravert you are I guess.

So this blog celebrates women who  ’just work in computing’ but who are slowly pushing back the boundaries so my daughter won’t have to.

I’m confused by the term “cloud communications”. At first I did a little research on Tropospheric Scatter, which involves bouncing radio waves off the lower atmosphere.

Wikipedia on Tropospheric Scatter

Then I realised that we were probably supposed to talk about ‘The Cloud’  as applied to telecommunications, but there again I have a problem. The phone system has been a distributed, network centric system of cooperating standards based elements transmitting user generated content for years! Wikipedia says the term “Cloud” was borrowed from Telephony, which makes for a rather circular definition.

Cloud Computing - again from Wikipedia

So in the end I decided that I’d take Humpty Dumpty’s line and:

“it means just what I choose it to mean — neither more nor less”.

Here’s my working definition: Cloud communications are the application of web-based cloud computing technologies to the communications businesses formerly dominated by the incumbent telcos. Principally voice telephony and short messages (no one has yet made significant money out of realtime video). In short, the next phase of the Telco (r)evolution.

There are some aspects of cloud computing (e.g Salesforce.com or Googledocs) that aren’t borrowed from the telco world, and the most interesting are :

1. Mobility and flexibility of service provision.
2. The ability of small organizations to assemble a user focused service utilising services from several large providers
3. Secure storage of data
4. Services that upload software to the user’s “terminal” to provide a customised user experience
5. Connection, location and device agnostic – the interface is the same irrespective of which computer and where I  login to a cloud service

The GSM mobile SIM ‘nearly’ provides the last of these, but when I swap my SIM into a new phone the rest of the context (contacts, recent call list etc) does not move with it. This leads me to wonder what the user ‘identifier’ will be for cloud communications. It won’t be a number- see my blog on the death of numbers , there are a few other possible candidates :

• SIM
• skype id (there are 400+ Million accounts !)
• facebook id
• google id
• twitter id
• role – in -company style ids – e.g. sales@phonefromhere.com
• SIP uri
• RFID tag (in a passport or bank card)

My guess is that two will survive, one will be ‘certified’ and used for commercial, legal and business calls, probably based on RFID, but possibly merged with SIM cards. (See here for my musings on this aspect).

The other surviver will be our ‘casual’ identifier, and I’d bet on Skype to win that one.

So why might telephony users move to new cloud based services? Social and business changes provide powerful drivers for people to migrate to services that meet their new needs better than the existing PSTN (or its VoIP clone).

• Decline in ratio of actual vs virtual travel
• Decline in physical vs virtual jobs
• Demise in Job-for-life
• Globalisation of trade and businesses

Some 40 years ago, my father (a diplomat) had a once a week phone call from West coast of the US to a small village in the UK. His call was exceptional. These days I often have two Skype calls a day from the USA and I’m no diplomat.

These drivers mean that people communicate with distant friends, colleagues and peers more frequently, but across more timezones over multiple media and with a greater need for clarity of understanding.

We (at PhoneFromHere.com) have been working with Family Systems on ‘VoiceChat’ which is a pilot cloud-based  voice system that addresses some of these needs.It is accessible (currently) via GoogleWave, skype, SIP, PSTN . The visual UI looks like this (at the moment)

Visual interface to Family Systems's VoiceChat

The important aspects of it are that it allows people to communicate synchronously in a ‘normal’ phonecall (via any of the above endpoints), each utterance is saved against the speaker’s name and timestamped. The participants to the call can later go back and listen to segments to refresh their memories and to annotate (and add audio comments to)  them. The conversation can then continue asynchronously  with folks contributing their comments over time, and including new voices that were not part of the original conversation. This adds wiki-like behaviour to a voice communications which works especially well across timezones, or even if you have a developer who works ‘different’ hours from the management team.

So what can we expect from the future of Cloud Communications ? As William Gibson (@GreatDismal) puts it:

“The future is already here – it is just unevenly distributed.”

Hints of what the future of Cloud Communications are to be found in -:

• Skype – I am where I log-in, on which ever device kinda find-me-follow-me without the fuss
• Blabbelon – push-to-talk online conversations in a stereo field
• Last.fm +  iTunes etc – internet audio on – the – go
• ZipDX  + HD audio + transcription (some people won’t participate in a podcast unless it is HD)
• Audio trails (around museums)
• AR – all focussed around images at the moment, audio is a much more natural interface
• Mechanical interface of iPhone + wii
• prevalence of iPods
• Second Life’s audio subsystem

We had tested OpenBTS and everything was working, so the next step was to mount the antenna on the local TV transmitter tower.

and install the OpenBTS unit into a rack in the hut at the base.

Then we switched it all on……

RESULT !!!!

Things started to go down hill at that point….

We realised that the Xorcom gateway which was to be used to connect OpenBTS to the island phone system was still stuck in customs in Auckland meaning that the system would have to be used in stand-alone mode for the time being.

We did some coverage and quality tests. We then found that the 900Mhz spectrum was pretty crowded – the old AMPS analog mobile system was present, which we knew about and had been warned to avoid by Niue Telecom. OpenBTS’s radio had been tuned to include a filter to ignore these frequencies.

We also found that there was widespread use of 900Mhz Wifi equipment – including on the TV tower we had our antenna on. This was a surprise, especially since 900Mhz is designated by the ITU as spectrum for licensed mobile communications in all parts of the world except North America. After some emails and asking around we found that the Wifi was run by the Internet Users Society of Niue who were using imported equipment that can be operated without a license in the USA.

It is worth noting that IUSN have had a complex relationship with the Government of Niue – see this article to get a flavour.

Anyhow, Harvind retuned the filter to try and block the wifi, which was threatening to overwhelm the signals from handsets. David adjusted the power management to encourage the handsets to put out more power (at the risk of poorer battery life) and we went out to test it.

We provisioned (by hand) some 30 test phones and started to use the system. Unfortunately an error on my part resulted in a Fijian visitor getting service, much to her and our surprise.

The system worked well, text messaging proving popular even in the small test group. Coverage however was more limited than we had hoped. We concluded that there were 3 reasons for this:

1. The vegetation in Niue is very lush – which we should have expected since part of the island is a rain forest. This dense greenery absorbs the GSM signal
2. The interference from the wifi was reducing our ability to detect weaker return signals from handsets (despite the filter)
3. The antenna mounting wasn’t optimal

We had a progress meeting with our client, the Govt of Niue and Niue Telecom. This resulted in a flurry of action. The islands fishery repair shop was drafted in to fabricate an improved antenna mount, which allowed us to correct the angle of the antenna, improving the coverage in the commercial center of Alofi.

Niue Telecom loaned us the use of a spare satellite link, which let us test international calling.

These changes improved the system to the point that it was very useful. Indeed on the final day I called into the VoIP Users Conference from my iPhone using the system.

Probably the best demonstration of the system occurred later that morning while I was doing some last minute shopping for souvenirs and my phone rang, (you should have seen the looks on the faces of the other folks in the shop!) it was our client, telling me that a TV crew had arrived at our rented house and could I get back quickly to be interviewed!

Overall the test was a qualified success, the main lessons learnt were :

1. Bring everything as personal luggage, ship nothing
2. Don’t assume a spectrum will be clear just because you have a license for it
3. OpenBTS works!

Links: Part 1 Part 2 OpenBTS blog

David makes last minute adjustments to OpenBTS

Within a few minutes of powering it up (and some DNS trickery) we had it working and made the first ever GSM call on Niue.

Tim places the first GSM call on Niue

Frank answers the call

Then we celebrated

Today we set up the big antenna check the coverage and start work on the provisioning system.

Links: Part 1 Part 3 OpenBTS blog

The island currently has no GSM operator – the only mobile phones are old analog “AMPS” devices. I can’t think of the last time I arrived somewhere and didn’t get a signal at the airport.

We are hoping to install OpenBTS to act as the GSM infrastructure on a TV tower, and connect it to the island’s telephone exchange using Asterisk . These open source components combined with some commodity hardware and some special purpose parts (antenna etc.) should combine to create a low cost mobile phone system for an entire (if small) nation.

Here’s what we have in mind:

As expected the unexpected has happened.

The only way to get to Niue is on the weekly flight from Auckland – we came on the plane last weekend, unfortunately our equipment didn’t. Much of it is still sitting in a customs warehouse in Auckland. We await the next plane with an-tici-pa-tion.

Fortunately we have quite a bit to be going on with. We have acquainted ourselves with the local conditions and started work on putting in the Wifi link from the tower to the telephone center.

I’ll keep you posted on how we get on.

Links: Part 2 Part 3 OpenBTS blog