This blog entry covers some of my biggest issues with the state of the telecoms industry as it stands today - but hopefully without ranting. Two things happened which started me down this particulr train of through - the release of the Ofcom report on "up to" broadband speeds and the Google Wave launch video.

For anyone that's missed it, its worth taking the time (its an hour long) to have a look at the Google Wave launch video; http://wave.google.com/. The code has been handed over to the Opensource community and I expect it to have a major impact on the way we use the Internet along with HTML5 - but its going to have impacts on so many areas from the network architecture up.

Its a part of my job at CBN to think about how networks will support people and the applications they want, and my first thought when I stopped thinking about what I could use it for, was that Wave and systems like it will bring back the whole debate around net neutrality, bringing into focus contention and what I used to call the "proximity to business" problem.

With Ofcom's "up to" broadband report still fresh in my head, it highlights how far old mindsets will have to travel to keep up with developers and users. Wave will, I'm sure, make use of big bandwidth but it would be just happy with less for a lot of what people will use it for; what it really needs is quality bandwidth which has low latency free, low jitter and with low contention - Ofcom pushed the the quality issues deep into their otherwise very useful report, fixing their gaze on the amount of bandwidth. Quantity is nothing if the quality isn't right also.

The rise of P2P and streaming reduces the viable contention level from the 50:1 that was sensible for ad hoc dial-up connections of old, past the 5:1 needed for on-net corporate networks, towards 1:1; gaming has the same impact. Real-time protocols like Wave will also push workable contention ratios from 50:1 to no more than 5:1 - not because they need gigabits of bandwidth but because they need reliable, deterministic bandwidth.

And this is the net-neutrality bit: Carriers that use traffic shaping and deep packet inspection to protect their core network interests rather than to consider customers will fail to deliver a reliable HTM5L/Wave environment.

This harks back to some of the work I did over 10 years ago, around the time Isenberg was talking about stupid networks. I never quite agreed with him - but never argued the point because what he said made so much more sense than the consensus at the time.

In my world, the edge of the network needed to be smart but the core needed to be respectfully dumb; third-ways were popular at the time. I liked the idea of applications and customer devices being able to tag their packets with information about the type of connection they needed - low latency for real-time, bulk transit for off-line applications, etc. The carrier edge networks could easily read these tags and set the quality of service to be respected by the dumb core.

Deep packet inspection has the potential to provide customers with a toolkit to see what traffic they are generating in terms of applications not octets and packet headers, and to allocate them into streams on the network. A decade ago I was routinely treating MIME types within Port 80 web traffic differently so embedded RealAudio streams were handled differently to general web, differently to HTML downloads. We rarely blocked traffic - we channelled and shaped it as streams started their journeys so there was room for everyone.

While we were working on massaging traffic, many of our peers were starting to block traffic; binary shaping with no nuances. The gravitational pull for all online applications towards port 80 was started in the '90s by dumb network managers who blocked everything they considered non-core activity. And so deep packet inspection was created out of myopia.

Even in those early days it was possible to see within Citrix thin client streams so on-line business applications could be treated differently to print jobs, ensuring network users recieved the optimal experience. Today its possible to reshape the edge of a network to any event an application developer can create or network manager can detect. You can dynamically reroute traffic onto different Internet transit connections based on how many prime numbers there are in the IP address; upgrade web performance because a customer logged-on to a paid-for application by monitoring a database table; increase latency because a satellite salesman signed up; anything.

Users who are dumb enough to cheat a system like this by dumping all traffic into the top grade stream find that every time they send an email their TV and phone will stop working - but no-one else will notice if its done well - but its their choice, not the carriers. This kind of stuff should be empowering not emasculating.

This technology is perfectly do-able today - its possible to do this on a £50 home router reflashed with Linux in a day or so. All the tools have been in the standard Linux distribution for 5 years or more. In fact, it feels like since I was working on this a decade ago, very little progress in using it possitively for the benefit of the customer has been made - the technology is stable but has been deployed facing the wrong way by people with no imagination.

Unless the network people start finding ways to become closer to users by talking to application developers, they will find that the next wave of broadband services won't meet customer expectations. In the work I did on smart edge networks I coined the phrase "proximity to business" as an issue for network architects, and the key reason this type of thinking would take a long time to be adopted. Its simply based on the idea that developers and analysts find this stuff easy because they have a direct relationship with customers (close proximity); server people find it more challenging because they have to use the developers as a proxy to the customer and their environment may be shared by multiple needs but they can typically be pooled to avoid unbridgeable tensions (throw another server together if needed); the network people, however, have a stack of people between them and the customer and their medium is necessarily shared by every competing resource.

So unless network architects make a conscious effort they have no viable proximity to real people and businesses. Without the concept of smart edge networks able to link to customers and the applications they use, network architects find they are too removed from customers.Their natural reaction is to protect what's nearest and dearest to them - the network, not people.

So far there has no evidence that carrier network architects ever really think about customers as people with needs - they rarely even think about them as "revenue generating units" that pay their salary - they seem to be irritations that damage the smooth running of their domain.

Tools like Wave will make the Internet too complex and demanding for this kind of mindset, and customers will get increasingly fed up unless networks people are brought into the business, rather than demanding that business continues revolve around them.