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

Friday, August 13, 2010

Work in Progress (comments welcome) - Code of Conduct for Policy Management & Net Neutrality

A food analogy: I see no reason why telcos shouldn't be allowed to offer "Internet-flavoured access", or "Processed Internet Substitute" if they wish, perhaps "Now fortified with Extra YouTube!".

But they shouldn't be able to sell or market those services as 100% Certified Prime Internet Access.

Like everyone in the telecoms industry, I encounter the issue of Net Neutrality and policy management very frequently. I've been watching developments recently, with the Google/Verizon announcement, consultations occurring in the UK and Europe, Chile deciding on a "hard" neutral policy, Canada choosing a middle ground and various other endless examples of posturing and lobbying.

It's got me thinking about what might be suitable compromises - a sort of generic "code of conduct" which might be applicable as a set of initial rules and goals, and which shouldn't irritate *too* many people. (Although clearly as we've seen with the Google/Verizon announcement, there's some irreconcilable differences between extreme ends of the spectrum of opinion).

I'm particularly thinking about rules for mobile data, which does have some practical constraints that do not exist in the fixed world. But given my recent work on fixed broadband business models and fibre, I'm also interested in more generic principles across the telecom industry.

The key areas I'm considering are:

  • Deliberate prioritisation, for example by guaranteeing QoS or providing an SLA for specific services, perhaps in terms of bandwidth, latency, jitter or other variables
  • Deliberate de-prioritisation, for example by blocking, throttling or otherwise interfering with a given data stream or service
  • Internet-based and non-Internet services
  • Internet access vs. other services
  • The problems of shared network resources being used simultaneously for Internet and non-Internet services

In general, it depends if there is congestion and contention for resources – if you prioritise something on an almost-empty network it obviously doesn’t impact anything else. Conversely, whether a telco should ever be allowed to constrain an application on an empty network is much more contentious.

My opinions (at this moment, and subject to debate and change) are:

  • Operators purely offering "in-house" managed services can do what they like in terms of managing their networks for non-Internet services - for example, on the core network it is fine to prioritise corporate VPNs, and on access lines they can tune their own IPTV or broadband TV services.
  • Where resources are used for both Internet access and non Internet services (eg home broadband lines, cell-site backhaul) and in particular where they are shared between multiple users, things get more complicated. In theory, it is OK with best-efforts "full Internet access" being at the bottom of the pile after other services have been dealt with. But there needs to be clarity in marketing, management and oversight. To use an analogy - if you buy a standby ticket for a flight, you know you might not get on. But if the plane takes off with empty seats and you're still refused boarding, you've got grounds for complaint & redress. Best-effort means that the service provider *really* has to make their "best effort" to accommodate you - *and* is still subject to the original contractual terms and subsequent reporting.
  • Internet access should be sold on the basis of average real-world speeds, not theoretical peak rates. Ideally, these averages (or perhaps minimum speeds) should be estimated for specific customers or locations (eg "given your house location, you should expect to get an average 2.7Mbit/s in normal use"). Providers should collect and publish actual average achieved rates, to allow comparison with their claims and promises. The stated averages in force at the start of a user's contract should not be allowed to fall during the course of the contract - this means that any subsequent sales of "managed" services on the same shared resources should be *incremental* to existing promised and contractual Internet capacity (calculated via statistical means)
  • The term "Internet access" needs some form of "Appellation contrôlée" - it can only be sold and branded as Internet access if it is not subject to undue policy management control, specifically around blocking/prioritisation based on application, flow, or destination. Anything else needs to be marketed and sold differently, eg "Internet-like service" or "Processed & reconstituted Internet, NOW WITH EXTRA YOUTUBE!". Consumer protection agencies need to be at the top of their game to ensure that end-users understand the difference - lessons from the food & drink industries should help.
  • *However*, it is fine for the provider to sell different classes of Internet access service, for example by speed, with caps or usage limits or other similar plan characteristic. An acceptable proxy for achieving application-level policy might be to have device-specific data plans, in the knowledge that (for example) BlackBerry users will typically behave differently to those dongles or tablets.
  • If capacity is available on the "Internet access" part of the network, it should be illegal to discriminate between different applications provided within that partition.
  • Any managed services (eg prioritised YouTube) should be provided by capacity in addition to that already dedicated to contracted Internet access. (Statistical estimations are fine for network dimensioning, as average speeds will be reported subsequently).
  • Renting broadband Internet access (fixed or mobile) should be similar to renting a property. The landlord should allow you "quiet enjoyment" without undue interference or limits, but is permitted to enter in emergencies, specify upfront rules (no pets) and hold you to account for destructive behaviour (withholding a deposit)
  • Governments and regulators may decide to mandate that all broadband providers give the option of "proper" Internet access, as well as any other managed services. This applies particularly in locations where it has been determined that Internet access is beneficial for economic or social reasons. It should be permissible to price pure Internet services at a higher level to managed Internet-like services.
  • “Negative” traffic management [throttling] on a per-application/per-service, on Internet access, is acceptable if (and only if) there are genuine resource constraints or costs, not just because an operator doesn’t like a given site/service
  • “Positive” traffic management is fine for non-Internet services (eg operator-hosted IPTV) provided over the same copper or radio as Internet access, as long as this is transparently described in detail to the customer
  • "Positive” traffic management [prioritisation] on a per-app basis, on Internet access is acceptable, as long as there are not unreasonable negative impacts on other apps/traffic. There must be absolute, detailed and realtime information about policy available to the end user and all application providers”
  • Negative / positive traffic management based on user, location, tariff, device type is generally OK, as long as it is not application/site specific [although these are often good proxies for applications]
  • Traffic management to manage integrity of network (eg DDoS attacks) or for legal requirements (regulatory prohibition of VoIP or services delivered to children) is fine
  • There needs to be an open “congestion API” provided so that telcos can be audited for fairness of traffic management policies
  • It should be illegal to block, prioritise or otherwise interfere with 3rd-party tools and services for measuring network quality, unless they themselves generate unreasonable traffic levels.
  • Operators must confirm that peering / transit partners apply equivalent rules
OK, I think that's a decent starting point. I'll hope to refine / iterate / clarify these over time. I suspect they will annoy both operators and hardcore Internet advocates. If I annoy both groups equally, I reckon I will have succeeded....


Todd Spraggins said...

This is the most reasonable and thorough analysis and proposal I have seen on the matter. I only wish more would listen to it (or would at least come to the same conclusions).

My largest concern over the issue has been the continual focus of neutrality in a technical context and not a business context. Having worked with or directly on these network systems, I have rarely found an engineer with malicious intent; they sincerely care about their networks and the quality of their service. However, history in this industry and in similar technology focused verticals, have proven that left unchecked, the technology becomes the proxy for commercial arbitrage, e.g. resetting P2P sharing of media by a competing media/ISP company. For this reason I'm still uncomfortable about commercial favoritism for a very specific "application" with-out knowing what protections others similar applications might have on the unfavored system. I have nothing against the ability to have differentiated express mail as long as I can still receive regular post at an acceptable level.

Because the internet is generative an offers so much potential, I think it is important that if such moves are made we still preserve the collective value of the internet in social policy. By this I mean OECD broadband counts would not count differentiated services toward economic indicators and especially in studies regarding digital have-nots. Furthermore, public providers should not be allowed to make these differentiated offerings when "basic" services are not competitively available.

Finally, I'm not thoroughly convinced that the "wireless is different" holds enough to go cart-blanch hands-off. If only transparency holds relevant, then is it completely acceptable for all providers to determine that video-calls are too costly to the network and bar them so that an entire market is transparently shut-out? I know these guys, having spent too many hours in 3GPP meetings and such- they are paranoid control freaks. Take femto for example. Why the hell are the major US carriers back-hauling this to the GGSN to finally offload. The transport distance and cost is more and my friends at Cisco have $$$ in their eyes as each of these users is a sub that gets counted in licensing fees! All because the carrier feels it has to touch the packet and "provide value". (LI is not an excuse as I have seen proposals to support such methods way earlier in the flow). Such behavior further chaps its users when you consider that both macro and femto count toward data usage caps. IMHO they have not proven that they can build cost effective networks; not because of technology constraints, but rather political and business.

(And speaking of data usage based billing, nothing outrages me more than having to pay for enormous video files for embedded ads, sometimes of the carrier's making, that go against your cap. Retailers are going to be in for a shock as consumer sentiment goes south over such ads and services sprout up to strip ads prior to going "over the air".)

Dean Bubley said...

Todd - many thanks for your comments, much appreciated and I agree with your sentiments entirely.

I'd suggest that you might want to make a clear distinction between The Internet (capital-I) and various other "internet" services. Then the OECD and others could account for them separately.

In a way, it's a shame that Internet isn't a recognised trademark....


Mo said...

I think you've almost nailed it, if I'm honest - it tallies quite closely with my own sketched out views on the subject.

Where you've said:

“Negative” traffic management [throttling] on a per-application/per-service, on Internet access, is acceptable if (and only if) there are genuine resource constraints or costs, not just because an operator doesn’t like a given site/service

I'd actually go a little further: all too often I see ISPs who throttle specific applications because they consume a lot of bandwidth. This tends to be a blunt inaccurate instrument, often with unforseen side-effects, especially where there isn't a well-defined port to filter on. The reality is that if a customer is using gobs more bandwidth than everybody they're sharing a link with and this is causing problems, then (a) you've got bigger problems than the impact of specific applications, and (b) it should be the connection you're applying measures to — after all, you don't see per-application throttling applied to transit for co-located servers (which, at some stage along the way, will end up on a shared link of some kind which needs to cope with the high levels of demand).

Thinking out loud somewhat, I wonder if there's some merit to advertising baseline vs burst throughputs on consumer connections, but that might be complicating things more than necessary.