Wednesday, August 27, 2008
Reading through various posts, my take is that the truth is a little more complex.
There are three main ways for developers/ITSPs to get VoIP software running on Symbian phones:
1) Use the native Nokia SIP stack & build your own application around it.
2) Use the native Nokia SIP stack and Nokia "Default" built-in Internet Telephony client, and build a much lighter application around these.
3) Build everything yourself, either with your own SIP stack or a proprietary mechanism.
Most wVoIP companies like Truphone have historically gone for options (1) or (2), although Fring is (I think) closer to number 3.
Lots of mobile conspiracy-theorists and bloggers are speculating that Nokia is somehow bending to the carriers' will by removing VoIP. I'm not to certain that's the main story. I think it's more likely that Nokia's decided that its default VoIP client is largely unused and fairly unloved, and in need of lots of expensive development and maintenance that isn't justified by customer demand.
Another possibility is hinted at on David Wood's (Symbian CTO) blog , which talks about competition vs. the iPhone. To me, this is a very telling comment:
"There are considerably fewer applications built into the iPhone than you can find in a standard S60 phone. That relative simplicity means that some feature-focused users will decide not to use the device. But the device taps into a new market that is arguably underserved by previous offerings. This is the very considerable market of users who don’t need every bell and whistle in feature-packed smartphones"
I'm wondering if Nokia is taking a hatchet to S60 overall, getting rid of some of the less-used features and applications that clutter its interface, in the hope of getting some more Apple-like useability. Pre-installed VoIP is certainly one thing I'd be thinking twice about in that context. If enthusiastic "featurists" want VoIP, they'll know how to get it. But for the masses, there's possibly more benefit in just getting rid of some of the confusing icons and menu items.
So it's unsurprising that mobile technology has also been dragged into the media whirl of the election.
Tomi has a great post about the use of SMS notification by Obama, although from a European perspective I can't help thinking that it's not exactly innovative to use text messages for political purposes. This is from Spain in 2004, and this from Hungary in 2002. It's also helped Morgan Tsvangirai in Zimbabwe.
Obviously theres loads of blogging, YouTubing and Twittering going on as well. But the latest fad in the US, apparently, is for campaign ringtones. And perhaps more interesting still is the use of candidate-specific video ringback tones by Vringo.
I'm not certain how well these will travel. I can certainly tell you right now that anyone I call that foists a video of Gordon Brown on me at the next UK elections, will be getting SMSs instead of calls until they switch it off.
Then I read the full article, which includes sensible commentary from Dan Warren saying "obviously you never get the top speed and they vary with distance from the base station and interference". Further, the GSMA's recent press release of 50m HSPA subscribers cites "peak data speeds over HSPA are currently between 3.6Mbps and 7.2Mbps. This translates to an end user speed of more than 1Mbps".
In other words, the GSMA is actually being realistic about real-world speeds for mobile broadband, and is in fact now trying not to fall into the trap of equating a cell sector's worth of shared capacity (under ideal radio conditions) with the achievable throughput of individual homes' dedicated wired connections. Good - HSPA is becoming hugely successful anyway, and doesn't need an overdose of cringe-worthy marketing hype.
I can't find any quote which directly suggests the GSMA thinks that it will beat fibre to 100Mbit/s. And in any case, talking about the UK in particular, it's probably worth suggesting that the GSMA might want to have a word with a couple of its members to help this along. O2 and T-Mobile UK currently have legal action pending against Ofcom which is delaying the 2.6GHz spectrum auction. Given the crowded state of the UK mobile market, that frequency band is pretty much the only one in which MNOs are likely to get the requisite 2 x 20MHz channels needed for LTE to run at 100Mbit/s. And it's also worth pointing out that even when that happens, 2.6GHz will struggle with indoor performance in many places, unless helped along by femtocells connected to - wait for it - fixed broadband.
Tuesday, August 26, 2008
The report included detailed analysis of what needs to change in handsets - both in the radio/protocol part of the phone, and in the OS/application layers.
My discussions with femto suppliers and operators recently have shown growing acceptance of the first part of this - that developments in 3GPP Release 8 specifications will need to be implemented in phones as well as networks. Most of the femtocell deployment models now seem to distinguish between supporting "legacy" pre-Rel8 devices, and the later, more capable ones.
Much of the work relates to the mechanisms for phones to register on specific femtos, avoid unnecessary attempts to register with unsuitable ones, and how certain femtos can be "preferred" - particularly the one in a given user's home or workplace.
The worst-case scenario is for a user to walk down a street with many homes with femtocells, and for the phone to attempt to connect to each one for the 5 seconds during which its signal appears strongest. This would kill the battery on the phone, generate a huge amount of signalling traffic and potentially lead to dropped calls. Various other scenarios are similarly unpalatable - for example, the phone not registering quickly with an expected femto (eg at home), and the user not receiving the promised discounted call rate.
The bottom line is that various new approaches need to be made to "white-list" particular femtos for particular devices, and aid the fast registration onto those cells. This will use a feature in Rel8 called CSG (closed subscriber group).
But while CSGs solve part of the problem, they also bring with them a whole raft of other issues. I discuss these in the report in some depth, but a key issue is how the user & operator can access and manage this white-list. Another issue is how a CSG-enabled phone can know when to look for a suitable CSG, without having to spend the whole time using battery power to scan for one.
Wading through various submissions to 3GPP, it strikes me that the majority of network-centric participants are not especially well-versed in the compromises that need to be made in phone design to accommodate these and other aspects. So you get lines in standards documents like "The user shall be able to request the UE to perform a scan for available CSG Identities." Well, that's great, apart from the minor question of "Yes, but how, exactly?" - presumably, this needs some form of connection manager application, a bit like using WiFi on a PC. Bear in mind that the average "manual network select" application on a phone is horribly primitive, rarely used, and usually buried 4 layers down in the menu structure.
As an operator, do you really want to be fielding customer support queries about provisioning guest access on femtos, on 100 different device and OS software types, with the user facing some clunky menu structure and indecipherable femto IDs that might look like WiFi SSIDs?
Based on both conversations and reading submitted material, the company that I see coming closest to "getting" both sides of the femto/handset problem at a radio level is Qualcomm. For example, this (zipped) document is a thoughtful analysis of the problems involved in finding femtos in various CSG scenarios. It looks at some of the issues around femto "public hotspots" and campus deployments as well as home-installed products. (It's also worth noting that Qualcomm invested in femto vendor ip.access recently).
One thing that Qualcomm is (corporately) very conscious of is the protracted timeline for getting any new handset technology into the market. As well as the network side of the industry, the phone and device chipset manufacturers need to get their heads around this issue ASAP as well.
On that note, I'll use this as a suitable opportunity to plug my Femtocell-Aware Handset report once again, and also that Disruptive Analysis is able to offer customised workshop and consultancy services on this and related topics. Please contact me via [information AT disruptive-analysis DOT com] if you're interested.
The latest in line is 3500MHz (and also 2300MHz), which until now has been the preserve of WiMAX and other similar fixed-wireless technologies. I suggested last November that this was a likely counter-strike by the invasion of the 2.5/2.6GHz band by WiMAX.
With the continued push towards a TDD profile for LTE, my view is that the WiMAX community is in danger of being further squeezed in some markets in terms of spectrum allocation.
The WiMAX Forum really needs to follow 3GPP's strategy and get profiles sorted for both FDD and TDD, in a broad range of frequency bands, as fast as possible. At the moment it is being extremely slow on this. It also needs to stimulate urgently the development of multi-band chipsets - it is probable that there will be only limited harmonisation at a global level, and even less on an operator-by-operator level. Frequency agility (and the ability to configure devices with different band permutations on the same silicon platform) will be critical for scale economies.
I've talked before about CS voice over HSPA data bearers. I've also thrown cold water on the notion of using IMS Multimedia Telephony (MMTel) as the main long-term VoIPo3G / VoIPo4G service.
So there's an important question outstanding... if the long term trend is to move mobile towards LTE (and maybe mobile WiMAX).... how do we deal with voice? Will all phones and networks need to support GSM/UMTS in parallel, in perpetuity? It now seems clear that not all LTE networks will have an IMS core. Some will, but many will not.
My view is that there are three main options in the long term:
1) Some good (and ideally standardised) ways to implement non-IMS NGN VoIP applications over LTE. This doesn't need to be a dumb pipe approach, but could use assorted hooks for QoS, security and so on, if well-engineered.
2) Use LTE as a pipe, and put Skype's / Truphone's / Cisco's / BroadSoft's / whoever's VoIP over the top of it. Deal with issues like handover to other technologies in software in the servers and the handsets - perhaps under control of the operators, but perhaps not.
3) Develop some way to "tunnel" traditional circuit voice over LTE bearers, as we're already seeing with CS over HS.
Martin Sauter over at WirelessMoves has done some sterling work looking at this problem (he calls it the Voice Gap) recently. He's just posted on a proposal for Option 3, as well a related proposal for what's being called IMS Centralised Services.
The idea of re-using CS voice instead of a shiny new VoIP application is not new. Many fixed-NGN operators' VoIP services "look" like circuit-switched telephony to the end user. Many ordinary PSTN services have long transited "virtual circuits" in an IP-based transport network. UMA-based dual-mode VoWLAN does the same - the handset's telephony application remains unaware that it's being transported over WiFi.
Obviously, this is bad news for suppliers of VoIP application servers, whether IMS-based or not. They need to push hard to demonstrate the added benefits of "native" VoIP as an application, as otherwise they risk missing a big opportunity around LTE.
Friday, August 22, 2008
And despite the supposed VoIP-blocking mechanisms, my friend Andy Abramson has cunningly found a workaround.
At one level I really don't like the idea of on-flight cellular/WiFi voice. There's enough noise on planes from screaming kids and tedious announcements about duty-free goods. But I certainly applaud another success against the ridiculous packet inspectors and their heavy-handed nannying at an application-by-application level.
Aircell should forget about blocking applications, and instead use their application-aware boxes to scrutinise what their customers want to use. And then improve and tailor the service to meet those needs.
Andy's idea to use Flash Audio makes a hige amount of sense. And I can think of at least two other ways in which VoIP could be disguised to get past censors like Aircell's.
Tuesday, August 19, 2008
About 200-300kbit/s seems to be a typical connection speed in the US, with some people reporting much worse performance (on the fringes of coverage). However, according to Wired's 3G iPhone performance map, many users in Europe are getting 1-2Mbit/s quite easily.
Various fingers of blame are being pointed - is it the AT&T network in the US? Is it the Infineon chipset?
Or is this just another manifestation of my favourite bugbear - the disconnect between mobile network designers and device developers (and thus by extension the 'real' user experience). It is simply the case that the original usage cases envisaged for HSDPA didn't include the type of rich, demanding applications (and implied traffic patterns) that iPhones generate?
- firstly, this could well be a manifestation of the HSDPA "idle mode" latency I discussed a few weeks ago. For iPhone users used to always-on, instant-connect WiFi, or even a nailed-up EDGE connection on a matured & optimised network - the initial "time to connect" could well be notably worse.
- secondly, in many countries, 3G is deployed in a higher frequency band than 2G (1900MHz vs 850MHz for AT&T, or 2100MHz vs 900/1800MHz in Europe). This means it will have shorter range, lesser coverage, and crucially worse indoor penetration.
- the audience of 3G iPhone users is fairly self-selecting: almost all actually use the data capabilities. While some of that is attributable to the phone's useability, it's also the case that it has attracted existing data-oriented users. It also tends to come bundled with data plans. This contrasts with most other popular 3G phones, for which only a small minority regularly use data (or even have a data plan). I'll bet the average Nokia N95 or SonyEricsson K-series user wouldn't notice lousy 3G signal, because they only fire up the browser once a month.
- various blogs have commented on the new 2.0.2 firmware release, wondering whether it contains changes to the radio stack. Some have claimed that they're seeing more bars of signal strength subsequently - although the cynic in me suspects it's easier to change the signal-strength indicator software, than the underlying radio.
- AT&T has not previously had the consumer 3G dongle phenomenon take off the same way it has in many other countries (reflecting different pricing strategies). So its network engineers may be a little behind the curve on dealing with massive, sudden ramp-ups of data traffic growth, often in new and unexpected geographic locations. They're probably faced with a whole range of optimisation headaches, and may even be needing to split cells & find new locations.
- Expectations of WiFi-like performance by end users reading about HSDPA's "headline" speeds may have been unrealistic. Normally, WiFi AP's only have 1-3 users attached simultaneously, whereas a 3G base station might have hundreds - with the available capacity in a sector shared amongst them all. Then there's another set of questions about the backhaul capacity from the cell site, in comparison with WiFi which usually has a home/office broadband connection to exploit.
- Somewhere there must be some side-by-side comparisons of an iPhone running next to another 3G handset (Nokia, Moto, whatever) connected to the same operator's network. If there was a big performance delta, that would point the finger of blame clearly at the phone/chipset rather than the network.
- It could be that the radio chip or antenna has worse performance on AT&T's 1900MHz band than on most European operators' 2100MHz for some reason.
- I'd imagine that the density of iPhone users in the US is higher than in most other countries, and thus more likely to put a strain on AT&T's network in dense urban areas.
Based on what I've been reading, I'm more inclined to point the finger at AT&T than at Infineon. Its 3G network has (to date) been geared more towards corporate PC + datacard users - and I suspect it's realising that massmarket consumer usage patterns are very different indeed.
Monday, August 18, 2008
Ofcom's just released a huge new report on the current state of the industry, incorporating telecoms, broadcasting and related services.
Some interesting insights I've spotted so far:
- Quite a lot of discussion of the resilience of fixed-line comms in the face of the mobile onslaught. Rather than direct fixed-mobile substitution, it appears that the UK sees more mobile-initiated incremental use of voice. Fixed minutes have dropped about 17bn minutes in total over 6 years, but mobile call volumes have risen by 38bn minutes. The UK outbound call total is still around 60/40 fixed:mobile, and 88% of homes still have a fixed line.
- The proportion of mobile-only households has been pretty static for the past few years, currently at 11%. This is considerably lower than elsewhere in Europe (eg 37% in Italy), and is possibly reflecting the prevalence of ADSL. Most mobile-only users are from lower socioeconomic groups.
- 44% of UK adults use SMS daily, against 36% using the Internet
- More than 100k+ new mobile broadband connections per month in the UK in H1 2008, with the rate of sign-up accelerating. 75% of dongle users are now using their mobile connection at home.
- Nearly half of adults with home broadband use WiFi
- 11% of UK mobile phone owners use the device to connect to the Internet, and 7% use it to send email. (It looks like the survey Ofcom commissioned didn't define "the Internet", so this might include some closed WAP usage too - consumers probably don't have a full view on what "the Internet" is at a technical level).
- VoIP usage appears to have fallen from 20% of consumers in late 2006, to 14% in early 2008. However, I suspect that this masks the fact that many instances of VoIP (eg BT's broadband circuit-replacement service, or corporate IP-PBXs), don't make it obvious to the user.
- Over two-thirds of mobile broadband users also have fixed-line broadband
- UK mobile subscribers send an average 67 SMS per month (or 82 / month per head, taking account of multiple subs-per-person). MMS use is only 0.37 messages per user per month.
- Slight increase in overall fixed-line subscriptions in 2007 - attributed to business lines.
- Overall UK non-SMS mobile data revenues were flat in 2007 vs 2006 at £1bn. I reckon that's because the data pre-dates the big rise in mobile dongle sales, and also reflects price pressures on things like ringtones. Ofcom also attributes this to adoption of flatrate data plans vs. pay-per-MB.
- UK prepay mobile ARPU has been flat at £9 / month for the last 4 years. That's a big issue for operators wanting to sell data services to prepay subs in my view.
- 17% of mobile subscriptions in the UK were on 3G at end-2007, although there's not much detail on the actual usage of 3G for non-voice applications.
- Overall, UK households allocate 3.3% of total spending to telecom services. That's been flat since 2003 - ie the slice of the pie isn't getting any bigger relative to food/rent/entertainment/travel etc.
- 94% of new mobile subscriptions are bundled with handsets.
- 11% of UK adults have >1 SIM card. Among 16-24yo users, this rises to 16%. There's an estimate that of the second devices in use in the UK, 1m are 3G dongles, 0.7m are BlackBerries or similar, and 8m are genuine "second handsets". There's also another 8m "barely active" devices that are used as backups, or legacy numbers that get occasional inbound calls or SMS
I'd imagine that at a global level, the figure is probably $20bn or so - and that's without the handset marketing budgets as well, plus retailers & assorted others. This contrasts with about $3bn predicted in new mobile-based ad cashflow into the industry, which various observers expect to rise about 5x over the next 4 years or so.
In other words, the mobile industry is likely to be a net spender on advertising until well into next decade.
It also throws up another conundrum - how much mobile-operator advertising will go via the mobile channel? Will you get a Vodafone banner while surfing on an Orange-supplied phone?
Saturday, August 16, 2008
(Caveat - these comments are all 2nd / 3rd-hand anecdotes, so some harder verification of this would be useful).
The key thing here is that Virgin is an MVNO on Sprint's CDMA network.
I'm guessing that the RF components of the Virgin phones, when in range of an Airave, go "Oooh, what a strong signal, let's attach to that base station!". But then something in the Sprint femto gateway double-checks the provisioning system and says "No. You haven't subscribed to the Airave package, so get lost". And then for some reason the phone can't "see" the other macro cells nearby (interference? something else to do with cell selection?), and disconnects entirely.
It's made me think about the more general problems of MVNOs + host-operator femtocells. This is not something I've heard discussed at any great length, probably because most MVNOs today are 2G, but most femtos are 3G. But going forward, it's going to add another layer of complexity to femto business models and possibly the supporting architecture. It may also prove to be another reason for future femto-optimised handset modifications that involve improved mechanisms for cell selection and mobility.
I've spent most of the time in the coastal town of Hammamet, which attracts a mix of local North African (Tunisian & Algerian) & European tourists. I also spent a couple of days in the capital Tunis and elsewhere. For those that haven't visited the country, it's sort of "mid-table" in terms of economy and wealth - roughly on a par with emerging parts of Europe like Macedonia or Albania, or Latin American states like Peru and Ecuador, and a bit higher than China's average (but without quite the same urban/rural polarity).
The first thing to note is that both my phones (one on O2 and the other on 3) failed to access data roaming at all. Ordinary GSM voice & SMS worked fine, but with both steered towards the Tunisiana network, I couldn't even get GPRS access or WAP to work. Maybe there was some fiddling-about I could have done with APNs on the phones, but frankly I really wasn't that bothered to make the effort. In any case, the networks are only 2G - although the shops are full of 3G phones, as people seem to like high-end devices for other reasons.
As in much of the developing world, the main business model seems to be separate purchase of unlocked phones and SIM cards / top-ups. Phones are sold in air-conditioned & modern retail outlets (I saw a full range from basic $40 devices up to $1000+ smartphones), but SIMs/topups are sold everywhere from tiny grocery shops to travel agencies.
Apparently I could have got a prepaid data SIM, but that seemed like too much enthusiasm (and a temptation to remain "connected" when I'm supposed to be relaxing). Instead I turned my phones off, except for about 10 mins a day to check any personal SMS's and voicemails, and also made a 10-minute trip to one of the Internet shops to check my email on a proper screen. If you never switch your phone off except for when you're flying, you should try it sometime for a couple of days - highly therapeutic!
Interestingly, I didn't see any evidence at all of local people using mobile data. No people obviously using handset browsers. Apart from the airport, I saw no BlackBerries. All the numerous mobile advertising hoardings I saw focused on voice and SMS promotions. The Internet shops had a fair number of locals (young and old), and I saw a lot of advertising for ADSL services. In Tunisia at least, the oft-repeated concept of "first usage of Internet on a mobile phone" seems to be clearly untrue at the moment.
Tuesday, August 05, 2008
Well, I see that a throttling detector has now been formally launched by the EFF.
I'm expecting Keynote-type live reports on ISP traffic-shaping tactics to emerge over the next year or two.
Eventually, it'll be a bit like a weather forecast "We're expecting some heavy VoIP blocking by ISP#1 this afternoon, with patches of throttled BitTorrent on ISPs #2-7 through the rest of tonight. The weekend looks pretty good though, with new neutrality laws from the US Congress and the European Commission coming into force on Friday evening".
Saturday, August 02, 2008
mobile phone cards [SIMs / subscriptions] in average". (It's on page 35 of the report, or page 36 of the PDF). In other words, there are about 450m unique individuals using mobile phones in China - considerably less than the 600m suggests.
What appears to be happening is that Internet use in China is being catalysed by wider availability of broadband, and more affordable PCs. 214m out of the 253m users are broadband-based. At the same time, there is some mobile use of the Internet - 73m users of the total access on phones - but virtually all of these are PC-Internet users as well. (There's no double-counting of mobile broadband as China doesn't have 3G yet). Looking at some of the charts on the CNNIC website, it looks like China's Internet use has hit a sudden point of inflection in the past 12 months, and is now on a steepening S-curve trajectory. Mobile is still growing extremely fast, but it doesn't seem to be accelerating at the same level.
This does not necessarily mean the same trends will be seen elsewhere in the developing world. It's worth noting that China is heavily pushing the rollout of fixed broadband - something which is much slower in markets like India and most of Africa. But it does suggest that China is extremely unlikely to have a future population of mobile-only Internet users.
One other interesting snippet from the CNNIC report - the average home Internet-connected PC has 2.7 users. This is worth remembering when considering all the stats on PC vs mobile handset shipments.