Speaking Engagements & Private Workshops - Get Dean Bubley to present or chair your event

Need an experienced, provocative & influential telecoms keynote speaker, moderator/chair or workshop facilitator?
To see recent presentations, and discuss Dean Bubley's appearance at a specific event, click here

Tuesday, September 08, 2009

Content: just a special sort of application

About 10 years ago, it became fashionable to say that "voice is just another sort of data on an IP network". VoIP, it was suggested, just turned telephony into mere bits, just like any others.

I want to extend and explore that description of subset/superset: I assert that "content is just a special sort of application on an IP network". It's just big chunks of software.

In both cases, there is a strong empirical evidence of truth, but the one word which is out of place in each sentence is "just", at least in the short term. Nevertheless, both assertions provide a view of how things change over much longer (10+ year) periods.

First, a quick re-cap of voice and VoIP. Certainly, at a network transport level, it is entirely feasible to get good voice quality with VoIP - most international calls use packet connections in the backbone. Millions use Skype and other VoIP services on the public Internet, while corporate IP-PBXs are mainstream. So at one, level, voice certainly can be viewed as just another form of IP traffic.

However, the word "just" is still not fully appropriate yet. Voice telephony has some very specific qualities that still mean it needs to be treated differently, and retains a quite high "value per bit". Aside from the obvious issues around latency and the need for realtime connectivity, it has some other , which have kept it at 60%+ of telecoms revenues. Firstly, numbering introduces barriers to entry and a need for interoperability. Secondly, regulatory issues remain wide and critical. Lastly, it has some very specific user-interface and experience issues - call setup times, the need for useable diallers and phonebooks, the natural human focus on time metrics (minutes) rather than IP packet volume, and so forth. There's also a hundred years of legacy experience, which conveys a lot of weight, authority - and sometimes arrogance.

The result is that voice telephony is indeed "special" - but over time, its importance and specialness are starting to wane. Basic voice revenues are falling - even in mobile, in some cases. The "phone call" is starting to blend into more complex voice interactions, new "voice 2.0" models are emerging in which context adds as much value as the media stream, and so forth. Networks are still designed with voice as a "special case", rather than just another packet stream. Voice is becoming blended with other applications and services, but arguably it is still over-represented and over-regulated compared with its long-term social and economic value.

Now, content. Compared with voice, the term "content" has always been a bit woolly in terms of definition; I've heard people refer to spreadsheets or even voice conversations as "content". But for most people in the industry, it tends to refer to visual programming media, chunks of written material, music, some images and so forth. The Wikipedia definition is:

"information and experiences that may provide value for an end-user/audience in specific contexts. Content may be delivered via any medium such as the internet, television, and audio CDs, as well as live events such as conferences and stage performances. The word is used to identify and quantify various divergent formats and genres of information as manageable value-adding components of media"

Like telephony, content tends to bring with it some specifics that do make it "special" - regulation, legal and commercial rules (censorship, copyright, reproduction rights & so forth). Video and audio content often requires special treatment because of their-sensitivity and huge volumes of data. And of course, there are decades (video) or even hundreds of years (text) of user experience. There is also a huge sense of entitlement by the media industry, that makes its advocates believe in their own uniqueness.

Yet for all its power and "specialness", it is starting to be put in its place. According to PWC/Cisco the entire market for digital media is worth about $300bn, including digital broadcast [non-Internet] TV. The software industry is also worth in excess of $300bn - and is shifting to either "cloud" applications or over-the-network downloaded apps like the iPhone. The video game industry alone is worth $30bn or more. Even the software piracy industry costs $50bn a year.

There's clearly many different ways to slice the statistics, depending on definitions or what's included/excluded - Internet vs. non-Internet, inclusion of things like SMS or web advertising, is user-generated material "content" in its conventional sense, and so on.

But to my eye, there's a close parallel here. Content is getting subsumed into applications like social networks or music-based communities. Amazon's valuation is about much more than the "content" stored in its warehouses or servers - it's the platform itself which is the core of the business. Facebook's value is about it's user base and APIs, not third-party chunks of media. Even for Apple, the AppStore is much sexier than iTunes, capturing a far greater share of industry attention.

And just like VoIP, digital content is also feeling the pinch of arbitrage on pricing. I don't just mean piracy - look instead to the failing of the print newspaper industry, as value moves to other application-based sources. (Are blogs "content"? I don't think of this post as an application, but I'd wince if someone called it content, in the same way I wince if they call me a blogger). And in future, what might happen to the value of news or live sports/music, if I stream and back-up all I see and hear to a server in the cloud via "life-streaming" or a similar application? Do content rights apply to my optic nerve?

Just as I think that the telecom industry is facing a dead-end in the notion of communication as simply sessions (see my recent posts on IMS), I think the media industry is similarly constrained in thinking of content as "chunks" of video or audio material. I also think that designing next-generation networks that are content-centric is as wrong as creating them session-centric.

Let's be honest. Ultimately, if voice is just data, rthen content is just software. We're not there yet, but they're both inevitable in the long term. Lobbyists and incumbents in both cases will plead for special treatment - justifiably, sometimes. Both telephony and content come with huge expectations on the part of users and regulators, and need to be protected in various ways.

But let's not lose sight of the end-game either - or entrench decades-old prejudices or business models in the underlying technology architectures. The content tail should not be allowed to wag the future application-networking dog.

15 comments:

Anonymous said...

This may be a stupid question but what on earth is a "content-centric" NGN? What is a "session-centric" NGN? What's the difference between the 2?

Anonymous said...

Maybe the question could be better phrased:

If I went to my favourite vendor and said "Dear Mr. Vendor, I would like to buy a content-centric NGN." What would he sell me? How would that differ from a session-centric Vendor's NGN? And how would both of these differ from just the common or garden NGN that we are all used to?

You pundits with your fancy makey-uppey language, worse than telcos I tell ya.

Dean Bubley said...

Content-centric NGN - various aspects, such as:

- NG access network built out first in areas chosen as demographically and competitively best for IPTV/triple-play
- Decisions on architecture like PON/P2P, FTTH vs FTTK based around business models very sensitive to TV revenues & uptake
- Heavy investment in mechanisms for prioritising / QoS-managing / billing video streams
- Economics based around set-top boxes rather than standalone modems/gateways
- OSS/BSS created with requirements skewed to enabling multi-party billing in scenarios where the user is being "sent" large chunks of content
- Assumption of ever-more asymmetry in traffic flows
- Heavy investment in "head end" stuff, cacheing, advert insertion, content ingestion & management & control, conditional access etc.
- May be less emphasis on aspects like IMS, more on control layers & subscriber data collection / analysis oriented around advertisers than the operators themselves.

Session-centric networks are rather more focused on investment in the control layer and core-network, rather than putting more emphasis on "the width of the pipe".

Economics based on assumption that continued large % of revenues comes from session-type services like voice telephony or SIP messaging, rather than HTTP, streaming etc.

I'm still thinking about exactly what features you'd want in an "application-centric" network, but a couple of guesses. It could include a lot of things like "edge-facing" interfacing so that end-points could make intelligent decisions about control.

For example, some way of flagging the current/anticipated state of network congestion would be useful for an app to decide what codec to use, whether to request QoS and so forth. Perhaps mechanisms for application-specific roaming or NNI as well.

It would an interesting exercise to ask a big set of developers what they'd like a network to be able to do, beyond simply being fast & reliable.

Anonymous said...

It seems that by "content-centric" you are basically implying IPTV and/or VoD support in a NGN with focus on FTTx for access. To my mind this is just "NGN" or perhaps NGA.

Session-centric NGN still confuses me, it sounds like what you are describing is a first-generation broadband network, not really what I consider to be NGN (nebulous as that concept is). Perhaps an example of an operator or 2 with this strategy would help?

As to the application control example you describe - TISPAN addresses your CAC example with RACF functionality although I don't know of any implementation of this yet.

Dean Bubley said...

I'd say that some of the municipal-led "open access" fibre build-outs are not especially IPTV-centric although clearly most are capable of it.

I'm thinking more in terms of the business case, rather than specific technology & architecture, although clearly that plays a role as well.

Same deal with the business case in the "session" scenario - ie there is an implicit assumption in the calculations for ROI etc, that much/most revenues will come from legacy subscriber-type services like voice plans. The notion that communication occurs in fixed, quantised "sessions" like calls is fundamental to both SS7 and newer architectures like IMS (which is based on SIP, ie Session Initial Protocol).

There are various other types of networks that don't have "subscribers" as such or an implicit assumption of "voice plans" or "content subscriptions", eg GSM-R for railways, assorted government and military networks, WiFi hotspots and so forth.

Anonymous said...

"I'm thinking more in terms of the business case, rather than specific technology & architecture, although clearly that plays a role as well. "

Plays a role? The architecture is fundamental. The business case drives the technology choice but the 2 are inextricably linked.

"The notion that communication occurs in fixed, quantised "sessions" like calls is fundamental to both SS7 and newer architectures like IMS (which is based on SIP, ie Session Initial Protocol)."

Are you implying that IMS (or SS7) drives a session-based business proposition? You have it wrong if so. It is the operator's billing system (BSS) that determines how the customer will be charged for a service(if at all). The IMS does not mandate any particular charging structure.

Dean Bubley said...

Anon> Plays a role? The architecture is fundamental. The business case drives the technology choice but the 2 are inextricably linked.

I'm confused. Are you the same Anonymous who was asking how I defined a content-centric network?

I replied that I defined a network as being "content centric" primarily by the business case used to justify its construction. I also added that the architecture has a role to play *in that definition* but it is secondary.

In any case, architecture is not necessarily fundamental: in some cases, any deficiencies can be worked-around at the application level to a good degree of satisfaction. This is becoming more true over time, as the endpoints (and the cloud) become smarter than the network.


Anon >Are you implying that IMS (or SS7) drives a session-based business proposition?

Yes, but I don't want to turn this comment thread of a re-hash of my IMS debate a week or so ago. IMS is optimised for those sorts of services that occur in clearly-defined "sessions". As indeed is SIP generally, unsurprisingly given the eponymous nature of the acronym.

IMS is not well-suited to business propositions that don't involve conventional notions of "subscribers", for example.

To give an absolutely random example off the top of my head, a network of a billion sensor nodes communicating both peer-to-peer and -to-server in realtime would perhaps not optimally use SIP/IMS. It probably *could* be done, but I doubt it would.

Same thing for an IPTV advertising service, that based decisions around subscriber data gathered from TV remote-control channel-flipping behaviour.

At a practical, ie commercial level, some proposed IMS services (eg anything using presence, such as RCS) are likely to be awkward in the context of prepaid mobile. Is the operator prepared to give presence "for free" when someone is out of credit?

Anonymous said...

"I'm confused. Are you the same Anonymous who was asking how I defined a content-centric network?"

Yes, I see where you are coming from but it's an entirely artificial distinction to my mind.

"In any case, architecture is not necessarily fundamental: in some cases, any deficiencies can be worked-around at the application level to a good degree of satisfaction."

Not true for IPTV, the architecture is key and deficiencies cannot be easily worked around.

"IMS is optimised for those sorts of services that occur in clearly-defined "sessions"

Your fixation with sessions is puzzling. The technical implementation of a service as session based has no bearing on the charging relationship with a customer - that is driven by the BSS. If there is a requirement for charging then it is only necessary that the architecture has some mechanism to record a chargable event, it is up to the BSS to then rate these and bill accordingly. This is true whether we're considering SS7 or NGN or VoIP or IMS or whatever.

Your example of a billion P2P sensors is a red herring. I don't see it as an example of something that would be billed on a per-session basis in any case. For example, if an electrical company were to contract out all of its power meters to a mobile operator for wireless connectivity, they would hardly be interested in charging a fee per unit of information sent. This does not preclude some IMS involvement though - perhaps such a service could make use of the authentication/identity features of IMS for example. Perhaps a burglar alarm service using networks of P2P sensors that are charged as one service to a mobile subscribers bill? I don't know, how would you expect such a service to be billed?

"IMS is not well-suited to business propositions that don't involve conventional notions of "subscribers", for example."

This is not a business proposition issue but more likely an implementation question. You seem to be under the mistaken impression that the industry believes IMS is a panacea. Nobody claims that it is suitable for all communication needs (noone is ever going to implement an email system on top of IMS for example).


"At a practical, ie commercial level, some proposed IMS services (eg anything using presence, such as RCS) are likely to be awkward in the context of prepaid mobile. Is the operator prepared to give presence "for free" when someone is out of credit?"

I don't see any practical difficulty with this example, it's a simple business decision.

Ericsson have invested heavily in IMS controlled IPTV, I'm not so sure about that either but they have demonstrated the concept at least so the notion of IMS controlling/influencing Ad insertion is perhaps not all that far-fetched.

Dean Bubley said...

> Not true for IPTV, the architecture is key and deficiencies cannot be easily worked around.

Agreed - I think video, especially HD displayed on a large living-room type screen, is very much a special case - both in terms of technical requirements and user expectations. It's an area I'm looking at for some new broadband business models, and certainly one where "content-centric" has some very specific technical implications. Although there are various possible approaches.

> Your fixation with sessions is puzzling. The technical implementation of a service as session based has no bearing on the charging relationship with a customer - that is driven by the BSS.

My issue with "sessions" has only limited relevance to charging at all. It is about the relevance to the way in which people (or things) communicate. People do not talk to each other in "sessions" or "events". It is a very poor "unit" for describing conversations or other interactions. For basic telecoms (ie phone calls) it's been a reasonable compromise as a lowest-common denominator, but that period is ending with the advent of IP. The charging aspect is secondary or tertiary. It's almost as arbitrary as using "syllable" as a fundamental unit of interaction.

> Your example of a billion P2P sensors is a red herring. I don't see it as an example of something that would be billed on a per-session basis in any cas

*Precisely*. Neither are many other forms of valuable communication - including speech in many contexts. So why have something "session based" at the core of NGN architecures at all? Why not just as an adjunct where needed (eg for things like authentication, as you mention).

> You seem to be under the mistaken impression that the industry believes IMS is a panacea.

No, my recent commentary has reflected on the notion that everyone thought it had gone away - in mobile at least. But recently, certain groups are trying to shoehorn it back in, not as a panacea but as a default.


All this discussion also moves it away from my original assertion: "content" isn't as important or unique as some believe.

Anonymous said...

"*Precisely*. Neither are many other forms of valuable communication - including speech in many contexts. So why have something "session based" at the core of NGN architecures at all? Why not just as an adjunct where needed (eg for things like authentication, as you mention)."

But why not have something session based in the core that can easily accomodate todays' (session based) services while also providing an interface for more sophisticated services? Why not choose an architecture that can accomodate all the scalability, legal and regulatory hurdles associated with large scale communication networks? Where is the realistic alternative?

The way I see it, some services will make use of IMS features, some will not, some will be billed, some will not. The IMS provides the flexibility to allow some services to use features as required. The fact that it records these as sessions is neither here nor there. It is up to the BSS to rate records from the IMS and bill (or not) as appropriate. The charging model need not be driven by the underlying technical implementation as you seem to be insisting (customers will not know or care that they are using IMS services).

Anonymous said...

@DB: People do not talk to each other in "sessions" or "events".

Technically, yes they do. A session may be a subset of the entire conversation or it maybe that one session is the whole conversation. In any case, in SIP signalling the session is just a technical convenience it has no bearing on how a customers services may be perceived.
It could be that an operator implements a service that requires no apparent set-up (e.g. IM) and it would appear to be session-less but the technical implementation can be based on SIP.

Dean Bubley said...

Anon> @DB: People do not talk to each other in "sessions" or "events".

Anon> Technically, yes they do. A session may be a subset of the entire conversation or it maybe that one session is the whole conversation.

I'm not a behavioural psychologist, so I might be on shaky ground here, but my impression is that normal human communication involves a complex & inter-related and non-linear dance of spoken words, body language, eye movement, intonation and a bunch of other contextual cues.

The traditional Person A to Person B phone call has been an extremely diluted form of this, slightly improved by adding video or hi-def audio.

Increasingly, extra nuances of context are being either reproduced or invented by new forms of communication. Social networks "add value" in unexpected ways by the ability to add in non-session components to an interpersonal relationship or communication. I'm thinking about things like voting, "like", new forms of sharing & invitation etc.

Sure, you *could* theoretically quantise these as "sessions", but if you think about this in the context of real life (micro-facial expressions, pheromones, touch etc) then I reckon it's nonscalable as it gets more sophisticated.

I think we are seeing Internet-based communications evolve in a parallel, complex fashion - for which linear standalone sessions are a very poor basic unit for describing interactions.

The problem is possibly that "sessions" are very convenient for billing purposes. But they are decreasingly convenient to ascribe chunks of "social value" in generic human communications.

As I said, I'm not a psychologist, but I meet very few people in the telecom industry who even start to think about Why and How people communicate.

Anonymous said...

@DB: "Sure, you *could* theoretically quantise these as "sessions", but if you think about this in the context of real life (micro-facial expressions, pheromones, touch etc) then I reckon it's nonscalable as it gets more sophisticated."

Why do you assume that an architecture that treats such individual units of communication as sessions can't scale? What sort of architecture would you suggest instead? Is this just an objection to SIP overhead? What protocol stack would you suggest is more appropriate?

Dean Bubley said...

> Why do you assume that an architecture that treats such individual units of communication as sessions can't scale? What sort of architecture would you suggest instead? Is this just an objection to SIP overhead? What protocol stack would you suggest is more appropriate?


Yes, in part SIP overhead is an issue. My understanding is that it is a substantial issue especially for multi-party or complex interactions.

A question for you: how easy is it to manage multiple SIP service providers, for users who don't want to funnel all their comms applications through a single gatekeeper operator? From work I did a few years ago, I seem to recall people referencing some issues with SIP routing on devices running multiple concurrent SIP applications, or worse, multiple SIP stacks.

I'd expect to see a heterogenous mix of protocols, much as we see today, with various optimisations for different scenarios. As someone points out further up the comment thread, nobody is going to reinvent email on SIP/IMS, nor something like a sensor network. Plenty of other things will use IAX or proprietary protocols, or stuff embedded in HTTP.

Mo Yan said...

About 10 years ago, it became fashionable to say that "voice is just another sort of data on an IP network". VoIP, it was suggested, just turned telephony into mere bits, just like any others.


Actually, tis :

"X" (insert your buzzword here) is
*basically* just another sort of
*SERVICE* , with *SPECIFIC* *CONNECTION*
and *TRAFFIC CONTRACT* requirements.

This has been obviously to me regarding
connections (network, sub-network
levels etc) since 1995 (the days of
SDH) , and regarding QoS (ditto) since
1997 (the days of ATM) + reinforced
in 1999 by UMTS.

I have a service S.
I come to network N.
I translate S into the connection
topology C (P-P, P-MP etc) , and the
traffic contract T (QoS etc) .


Then I do : N.give-me-service(C,T)

Mobility is merely a parameter of T.
That is what 4G is all about.