The Number 1 problem for "seamless" models of WiFi offload / network selection

I frequently argue against the assumption that "seamless" connection to WiFi is practical or desirable, especially for the general case of smartphones or tablets being "pushed" automatically to connect via a combination of technologies and new standards like Passpoint, Hotspot 2.0, ANDSF and EAP-SIM.

I thought it was worth stressing and clarifying the single most problematic use-case:

Worst case scenario: A device automatically being connected to paid or overly policy-managed carrier-WiFi, when a free alternative is readily available

Nothing will cause customer dissatisfaction & churn faster than incurring additional costs (or additional usage against a quota) than this. Expect lawyers, regulators, consumer press & commentators to throw very large & painful bricks at you if it occurs. And probably device & OS vendors too.

The classic example that most readers of this blog will be familiar with is attending a conference in a hotel with expensive WiFi, but where the event organisers give you a free code, or even set up their own APs for better performance/lower cost. We'd be deeply annoyed if our phones "roamed onto" the local WiFi automatically, unless the usage was 100% free as well. Similar story if your frequent-flyer lounge has free WiFi, but your devices magically jumped onto the paid/roamed WiFi in the main airport hall.

I am writing this on a laptop, with two phones by my side, in Starbucks in London. There is free WiFi (provided by BT Openzone in this case). You typically log on via a very simple splash page and selecting an obvious Openzone-Starbucks SSID. A very small "seam" indeed. Alternatively, I can use the BT WiFi app I have on my phones for free, or login on the "normal" BT Openzone SSID page, because I have BT Broadband at home. I'd be irritated if I was automatically shunted to one of my CellCo's SSIDs instead, with their charging and policy being applied.

Same thing in hundreds of other situations too - restaurants, pubs, offices, friends' homes, public buildings and so forth. There is free WiFi - often very good too - easily obtainable from a source other than the cellular operator. There are multiple stakeholders here - the venue, advertisers, content companies, fixed operators, your device supply, your employer and so forth. All might have negotiated a better "private" WiFi experience than your cellular provider at that location.

Not only that, but the cellular WiFi option might be subject to onerous policy restrictions (eg VoIP blocking) which don't apply when you use WiFi in "private" mode. 

There must be an easy and obvious way to manage this - and not just buried 7 layers down in the menus with some obscure configuration settings. The alternative is that the carrier WiFi is 100% free, 100% unmetered, in which case people won't care about the other options. That is why Apple is happy with AT&T and some other operators doing automated WiFi log-on, as it's in a customer-friendly form. One other possible option is that the operator (instead of charging you) gives you extra free stuff such as exclusive content to tempt you onto its WiFi option.

Seams are often there for good reasons, as well as bad. Seams can be monetised; seams can present the user with important information or choices. In some cases automation and "frictionless" connection in desirable, while in other cases it's not. Mobile operators are just one (often minor) stakeholder in the overall WiFi ecosystem, and need to stop arrogantly assuming that their WiFi choices trump either users', or other parties. Unless and until Hotspot 2.0 and its related standards fully and explicitly recognise these choices, it's dead in the water.

WebRTC - how to get support into old browsers?

I'm currently updating the WebRTC forecast model used in the Disruptive Analysis research strategy report & update subscription service. A core part of it examines the likely penetration of WebRTC capabilities into different classes of device, via browsers or other paths.

Most in the industry will know that Chrome browsers already support WebRTC switched-on by default (albeit in its current pre-standard form), with Firefox support close behind in its Beta version. There is much speculation about if/when Apple and Microsoft might follow. I've got my own views on that, with my assumptions stated and baked into my forecasts.

But there is a critical difference here - there is a very large legacy installed base of IE browsers, and to some extent Safari as well - that doesn't auto-update to the latest version. IE9, 8, 7 and even 6 are still lurking around, especially in certain countries either with lots of old PCs, or for corporate users whose CIOs lock down application updates and installs.

How is WebRTC going to be extended to those last hold-outs, even if IE11/12 or future versions of Safari support it?

One option commonly discussed is Google's Chrome Frame plug-in for IE. However, it is unclear whether anybody who still uses an old version of IE is going to readily want - or be allowed - to install a new plug-in. The whole point of WebRTC is to avoid plug-ins, especially for occasional audio/video tasks such as conferencing.

(Sidenote: I have a growing hatred of all the various web conference tools that vendors briefings or clients/contacts force me to use - no, I don't want to have to sign in 15 minutes in advance to make sure it all works and I've downloaded whatever chunks of software are supposed to help. Increasingly I turn down such requests, asking instead for a PSTN dial-in and emailing me the PDF presentation - or, alternatively, force *them* to use my choice: Skype).

So, I'm doubtful that Chrome Frame is going to do much to extend the reach of WebRTC to many of the old-browser community. Maybe Google can cunningly force its use by pushing it as a way to enable/enhance GMail or Hangouts, but it's not obvious that's the case.

I had an interesting thought yesterday though - most of those old browsers will already have Flash and Java plug-ins (as well as maybe Silverlight and others). Maybe the "vector" for WebRTC support on "old IE" will turn out to be Adobe or Sun (ie now Oracle), if they include it within the next updates of users' current plug-ins? Obviously Adobe still has its own competing RTMP & RTFMP protocols, but it must surely by now see the writing on the wall, much as it has with Flash vs. H.264? Could it gain relevance by being a lead provider of WebRTC APIs? Provide a combination of the two with RTFMP as a fallback or way to manage certain scenarios? Maybe even get subsidised by Google to help extend WebRTC's reach more quickly....

The Oracle/Sun/Java pitch is perhaps easier - it wouldn't surprise me at all, and there are some obvious synergies with the acquisition of Acme Packet.

(Note - I have no inside knowledge of either of these; both are total speculation on my part. I'm actually embarrassed I didn't ask Oracle or Acme about it - or even just suggest it - when I met them recently).

One question I haven't thought through though - what would this mean for PCs which end up with two or more WebRTC implementations? Native browser capability, plus plug-in, plus maybe the sort of web-app Javascript/API approach we see from AddLive or Plivo and others? Will there just be a "default WebRTC engine" or will it be managed in another way? It strikes me that we'll definitely see smartphones and tablets with 2+ implementations of WebRTC, so this issue will need to be addressed anyway.

Telcos need to offer 2+ separate voice applications & decouple RCS + VoLTE

In yesterday's post about RCS, I mentioned in passing that VoLTE and RCS must be 100% separated if either is to stand a chance of success. At present, attempts to combine them in RCS5  are, depending on your viewpoint, either: (a) a rather Machiavellian attempt to force deployment of RCS even among skeptical carriers, or (b) because a few use-cases of IM/chat benefit from being integrated with voice communications. There is also the vendor bundling argument of "Get 2 apps for the price of 1 on our IMS platform".

The net effect is shackling the already-hobbling VoLTE to the corpse of RCS, making it drag its deceased counterpart along for the entire distance of a marathon. Not a strategy for success, for either (a) or (b) scnearios. To mix my metaphors - just because you share IMS DNA with your sibling, you should still remember that incestuous marriage is banned for good reasons: your offspring will likely have undesirable genetic traits. Far better to look for fresh genes elsewhere and cross-breed with Internet species, for example.

That led to me to a broader set of thoughts about voice/messaging integration, and also fragmentation of the voice experience. I've often used this chart (or variations) for the past few years, and it encapsulates what is happening as voice goes "beyond telephony":

In a nutshell, voice communications is changing from standalone telephony, to a more complex and dynamic market in which voice gets embedded into many applications and contexts, often in forms that don't look like traditional "phone calls". WebRTC accelerates this, although it is already happening via apps and APIs in any case.

This also means that typical phones and PCs will have multiple "voice apps" in future, besides the dialler. Obviously this is already becoming true for many users today, who have VoIP capabilities from Skype, Google, Telco-OTT apps, corporate UC systems, embedded voice in games or Facebook and so forth. With the advent of WebRTC we'll see even more fragmentation of the voice experience, with websites and mobile apps featuring voice functions internally. Often, the experience of voice will often be nothing like a phone conversation - no "Hello, is that Dean?" or the normal interruptions, ringing, stilted etiquette and other bits of 130-year old furniture that make up this thing we recognise as a "call".

While we already see cloud providers and operators provide telephony APIs to developers (some do already), many of those still have the same "raw ingredient" as the basic phone call, using the same voice engine and numbering as the handset's main dialler. That's not good enough.

There needs to be much greater flexibility and imagination. For example, if I'm calling from a call-centre there needs to be Grade-A background noise suppression so you can't hear the inane chatter of my colleagues, while I try to sell you double-glazing. But if I'm on a beach making a "wish you were here!" call to a friend, it would be nice to have the gentle sound of waves in the background, not deathly silence. There are many other parameters and interaction models for developers to tune too - that's just a simple example.

For video, there are actually more options, because firms like AddLive, Plivo, Telefonica/TokBox and others are offering video APIs and back-end cloud services, based either on WebRTC or proprietary alternatives. But as far as I know, there aren't many easy APIs for non-telephony forms of voice communication yet, although Twilio, Rebtel, Voxeo Labs and others enable in-app voice with some control over interaction logic.

Coming back to the RCS/VoLTE debacle, this points to an uncomfortable conclusion - rigid, standard phone calls are precisely the wrong sort of voice to include in RCS-type use cases anyway. Even if RCS miraculously became a success, developers would probably want to mash it up with non-VoLTE voice a lot of the time. You'd want stereo for a karaoke app, and the capability to deal with a lousy external microphone, for example. Some might want to trigger a traditional phone call, but others would not. Even then, developers will want a choice of 3rd-party telephony/VoIP APIs and not just the carrier's, for example if they want to use different phone numbers.

So if telcos and their vendors decide to put voice capabilities into RCS anyway, they need to start from square one, and think "Why are we putting voice here? Is it for phone-type calls similar to Skype? or is it for other use-cases? What do users and developers actually want to do?". As with my post on messaging, they need to think about Intent & Purpose, and not just get blind-sided because they can get both sets of (independent) capability from the same platform.

A similar argument applies to VoLTE adding RCS. If LTE phone calls can benefit from associated messaging, presence or other features, is RCS really the right/only product to deliver that, for the specific use-cases that carrier feels are important?

Operators should be looking more at mashup / orchestrated approaches. And they need to be happy to have multiple voice experiences on their customers devices - and not only that, but they ought to be developing multiple voice apps/services/APIs, not just clunky old telephony warmed-over in VoLTE guise.

So if you want voice in RCS, then there's a big choice out there. Bolting it to VoLTE makes no sense whatsoever - it would even be better to have a second, different VoIP app-server on the same IMS and use that instead, especially as half the time it's likely to be running over 3G or 3rd-party uncontrolled WiFi anyway.

An RCS use-case I'd actually support & see as viable

It's now more than 2.5 years since my publication of a report on "RCS is Dead" in September 2010. While RCS has shambled on as an undead Zombie technology since then, and we've still got a few lonely operators kicking over its rotting corpse at the moment, I thought I should reflect on a couple of points I'd mentioned.

I maintain my position that there is zero benefit - and quite a lot of cost - to RCS (either the general version or the Joyn-branded GSMA-approved one) as a standalone app. The contention that it will somehow encourage people to give up using WhatsApp, KakaoTalk, Line and the other messaging platforms is risible. Worse, expounding the notion that RCS might sway people away from using embedded messaging capability in apps such as Facebook should be grounds for firing someone for gross incompetence. Blending it in with VoLTE will further delay and complicate an already-struggling (and much more important) system - the two should be kept 100% separate until either/both are well-established.

In my 2010 report, I suggested that one way to salvage something from the wreckage of RCS was for operators to offer fully-OTT versions of the app or API. MetroPCS is fairly close to that, announcing a related capability at MWC . I  also suggested that operators consider using RCS for their own customer-service and self-care function, and mandating their own employees to use it for IM (not eating your own dogfood is never a good sign). A fourth option was using RCS within operator-branded vertical-community apps, perhaps for sports or movies or home-automation.

In many ways, the GSMA Joyn branding, which has to be paid for (!!!!) by operators wishing to subjugate any hope of differentiation, has made matters a lot worse. It just highlights the fact that it is a clumsy multi-operator service that has little value until 80%+ of a given population can use it. (For MetroPCS, it's perhaps not so bad as it's the only RCS player of any sort in North America at present, so it has the logo to itself for now at least).

Overall, I see no grounds for a general-purpose, cross-operator standardised messaging app. Pointing to SMS as a success "proving" federated standards are worthwhile, ignores 20 years of innovation on the web, smartphones, cloud and apps in the meantime. Standalone Joyn is worse than useless, sucking up money and resources that could be better used by operators on design and innovation, or partnering with Internet/app providers who "get it".

I've got more sympathy - to a limited extent - for RCS as an API, or with various forms of web mashup which might create new messaging experiences and user-interaction models. Because the web integration would likely be single-carrier, ideally offered on an OTT basis rather than federated across multiple telcos, there is a chance to differentiate and make some money. (I'm not going to cover WebRTC integration here, which I address that in quite a lot of depth in my recent report ).

But in the spirit of my original report, I've actually thought through a variant of RCS which I'd actually be prepared to support, and even use myself if it worked well. Yes, you read that correctly.

For *some* of my messaging interactions, it would be useful to have a common-denominator app that's better than SMS and more convenient than email. I've thought of a version I'd actually be prepared to use.

I hate unsolicited SMS and spam emails with a vengeance. I want any future messaging services to be white-list capable, with sophisticated controls and policies. I'm not convinced that Facebook or Twitter - if they ever released an open messaging API - would be in a position to control their advertisers to the extent I'd like.

So, thinking about my "perfect generic messaging app", I want something which is able to:

- accept inbound messages from my Facebook & LinkedIn contacts only (including groups & events, not just people, but excluding brand pages). I want that configurable so I can add/drop other social services - and subsets of affiliations within them - in the future.
- accept inbound messages from people in my handset addressbook (these days, that's far fewer than FB & LI contacts - I rarely take peoples' phone numbers).
- charge £5 a time for inbound messages from unknown people or organisations, based on a combination of number + social connections. I'll even give a rev-share on this! (There's your new business model, telcos, protecting my privacy and acting as my "interruption agency").
- give me the ability to refund the fee if I want (eg a friend loses their phone & contacts me from someone else's, or a prospective client gets in touch)
- have a "report as spam" and "block" button, which also (& quite specifically) will block all marketing messages from my service provider if I choose
- has a sensible price structure - maybe not quite as cheap as WhatsApp, but £5-10 a year flat is reasonable, excluding the value-adds like the pay-to-interrupt feature. No international or roaming premiums
- has a good UI and features that evolve on a monthly basis
- Service is offered on a fully-OTT basis, usable from all my devices, irrespective of connection method. I want it to work on my PC without installing an application, via the web with a password login (or FB / LI / Twitter authentication)
- Decoupled from SIM and mobile phone number, using a separate identifier range, so there is no need to "port" it if I switch access providers, and it works easily when I travel and use a local SIM in a spare phone
- Integration with my preferred voice applications (eg click-to-speak/call), especially Skype at the moment, but maybe WebRTC ones in future too.
- A good / clever UI. Facebook's new "chat-heads" is fantastic
- Doesn't do anything creepy in terms of "big data" through analysis of my social graph / usage patterns
- EDIT: Doesn't kill my battery with clunky presence / always-on implementations. (Hat-tip to Kevin Holley on Twitter for that!) Whatsapp's"last seen online" is probably the best current version of presence, rather than online/offline/busy which nobody uses properly.

Anyone prepared to step up for that? Is it easier to do with RCS, or does that add little to the solution? For me, the "no unwanted interruptions" feature is critical, as is "report as spam".

Standalone, and pitched as some unconvincing "SMS 2.0" or direct clone of Whatsapp, RCS & Joyn are worse than useless. But if you can use it to create something unique that improves the user-interaction model, then I'm willing to listen.

(Sidenote: in the past 6 months, my worst SMS spam offenders have been the UK's pestilential PPI-refund lawyers, my own operator Vodafone UK, and the GSMA sending me unsolicited nonsense about MWC).

There is no "messaging market"

I keep reading ridiculous articles talking about SMS, so-called OTT applications like WhatsApp & KakaoTalk, Facebook, and of course my favourite RCS/RCSe/Joyn family of services.

A central theme in these articles is a supposed battle for the "messaging market", with lazy journalists or vendor marketeers painting a dark picture of mortal combat between the righteous fortress of SMS revenues, the marauding hordes of barbarian OTT players at the gates, and the Knights of Joyn riding to the rescue in their shining armour of interoperability.

This is all palpable nonsense - a strawman argument to reframe a complex and dynamic situation into the usual fatuous and imaginary Us vs. Them, Telcos vs OTTs narrative, coupled to a desperate attempt to make RCS and its GSMA-branded offspring look relevant.

Not only is this argument flawed, the likely outcomes will in many cases be worse than useless.

There is no "messaging market" today, and there certainly won't be tomorrow. We have multiple different mechanisms to send "messages" across the Internet and between phones today - and more to the point, many, many reasons and contexts for wanting to do so.

The idea that a "he said, she said" gossip via SMS is in any way equivalent to sending a signed NDA via email is ridiculous. Almost as ridiculous as equating a WhatsApp chat between friends, to a online-customer support IM session embedded in a software company's web-page.

You might as well say that there's a market for "sending things coloured blue", as it makes no less sense than a market for "sending messages". It's just more difficult to count.

For most of the last 15 years, the bulk of messages have been sent by SMS (mobile-to-mobile or server-to-mobile), Internet email (PC-to-PC, server-to-PC, or PC-to-mobile) and Internet IM (mostly PC-to-PC). And because emails are generally longer and "richer" than SMS, nobody has bothered to talk about the risks of messaging substitution between them.

As Internet access and apps have appeared on mobile phones, unsurprisingly we've seen an upsurge of both email and IM on devices. We've also seen a massive growth of other ways of sending messages such as push notifications, and especially in-app messaging for social, enterprise or gaming purposes. Quite often, we'll get the same message - or a notification - sent through multiple channels simultaneously, or to multiple devices, so there is rampant double-counting too.

There are actually three- or perhaps more - completely separate trends occurring:

1) SMS is losing its crown as the "standalone" mobile short-message platform of choice, because it is overpriced by around 100x, and hasn't evolved in feature-set in 20 years.
2) Standalone messaging is starting to diminish as it becomes easier and better to send contextualised messages
3) People only need any-to-any interoperable messaging when there's nothing better available. Most of the time, they're contacting friends or colleagues who'll have whatever app of choice works best in that group.

The key thing here is "purpose". SMS and email are (to a fair degree) general-purpose tools. Jacks of all trades and masters of none. An SMS can be a notification of a delayed flight, directions to meet your friend in the pub, a spam advert, or a way to send configurations & settings to your phone. Emails are similarly diverse.

General-purpose tools are fine when specialist tools are unavailable, expensive or inconvenient.

Do you think anybody would use a Swiss-Army Knife to open a bottle or cut a log, if the person had a magic app on their phone, or in their browser, that could instantly manifest a proper corkscrew or saw, at zero extra cost?

Moreover, would anyone other than Victorinox & Leatherman care about the demise of the pocket multi-tool market? Would they be stupid enough to try to create a new "Rich Tool Suite" standard between them, with interoperable and modular scissors and blades that could fit into a new generic multi-purpose tool body? And would industry analysts start adding together manifestations of tweezers and hacksaws to try to draw numerical comparisons with the legacy multi-tool market?

Just as I've previously mentioned that there is no "video" market, and increasingly no single "voice" market, there is already a clear fragmentation in messaging as well.

Yes, for any given use-case there are overlapping Venn diagrams which demonstrate substitution. I can tell my friends which pub I'm in using SMS, Facebook chat, WhatsApp or Twitter DMs if I wanted to. I could send my signed & scanned NDA to a client via email, Dropbox - or even fax or by mail.

Adding together some arbitrary subset of messaging tools - SMS + OTT apps, for example - but missing out push-notifications, emails, HTTP in-browser IM and so forth AND neglecting to de-duplicate where single messages...

... are cut into...

... chunks for convenience & readabilty...

or where the same message is sent & notified by multiple paths, just means that the data is spurious and the analysis sloppy. You never see any segmentation of messaging by purpose, except perhaps for person-to-person vs. app-to-person.

Any worthwhile analysis would look at various ways to slice up this supposed monolithic market into separate buckets reflecting context or intent. Perhaps social messaging vs. advertising vs. standalone information vs. gossip vs. B2B meeting arrangement vs. one-way app updates. Or sliced by length of a messaging "session" or number of participants, or a hundred other ways.

That type of analysis would also point out why the rhetoric around RCS and Joyn is a mistake. The market for general-purpose messaging is fragmented, declining in relevance, and simply not going to be very valuable in the future. WhatsApp's 1$-per-user-per-year price point is a good indicator of where it's going. Operators ought to be segmenting the messaging space by intent and developing laser-like tools where they can differentiate, not wasting time and resources on creating a lowest-common denominator Rich Tool Suite with the added costs and politics of IMS and vendor hype.

Moreover, the random selection of "capabilities" in RCS is the same set of pointless ideas for services that have been around for at least 10 years. "See what I see", "File Transfer", and old-fashioned and useless forms of presence? That's Yahoo Messenger from 2003, not a modern mobile communications app - see Vine's video clips, or Facebook's chat-heads, or WhatsApp's "last seen online" for some ideas about application *design* rather than engineering.

The irony is that the telecoms industry has been greedily absorbing $100bn+ a year from SMS and MMS, yet has plowed back almost zero into R&D and messaging application design. No sympathy or regulatory recourse whatsoever is due for telcos, as that market collapses - it has been a dead man walking for years. It has been neglected by companies asleep at the wheel and unwilling to innovate - there isn't even an easy "report as spam" function, or a way to switch to whitelist-only.

To be honest, I think WhatsApp and its peers will face challenges too - not only is it targetting a standalone messaging market that will implode in value, but it's doing so with constraints like using phone numbers as identifiers. That's already looking clunky in a multi-device user world, especially where many are WiFi-only.

Just as with voice communications, it is wrong to think of messaging as a meaningful and homogeneous "market". In most cases, messages will be sent as a feature or function of something else, not as a "service". There will be diminishing need - and diminishing value - in standalone, general-purpose applications. The true value will lie in many separate areas and business models, reflecting the myriad reasons why people or machines want to send messages in the first place.

So stop talking about the "messaging market", stop counting messages and drawing faulty conclusions from easy/lazy-to-quantify numbers, and start thinking about Purpose, because that's where the value lies.