By the mid-2000s the minimal cost-to-serve a mobile phone user had got down to the point where it was worth Roshan's while to put base stations in places where British soldiers broke down 105mm light guns to carry them piece by piece up a cliff, in order to fire from the hilltop next to the base station and get additional range.
It's fairly well known that the Taliban weren't entirely pleased about this, especially when ISAF started publicising their tip-off hotline and people did just that with their new second-hand Nokias. And they started destroying base stations until the operators agreed to shut down for part of the day. An uneasy settlement was arrived at - after all, Talibs use the phone too, and so do their families and friends. Like the old pattern of the insurgent owning the roads during the night and the government during the day, the insurgent owned the 900MHz band during the night and left it to the government during the day.
(However, their control of radio spectrum is purely negative, as if they were to use it themselves, the government could spy on them doing so, direction-find the transmitters, traffic-analyse the network to find out who is important, and sic drones, attack helicopters, or commandos on them. They can intimidate other people out of using it, but they can't use it themselves without very careful security precautions.)
So I'd like to recommend this really excellent article.
It seems that this shaky modus vivendi has broken down. Not only are the Taliban destroying more sites, they are doing so more thoroughly.
A typical problem for an emerging-market GSM operations engineer is the security of diesel fuel. Some operators in Africa are their countries' biggest electricity generators. This is fiendishly expensive - not only do you have to buy the diesel, you have to pay people to fill up the tanks on thousands of remote cell sites. And other people will steal it, or even steal the whole generator, which is why some of them are half-way up the tower although that means the structure must be much heavier and stronger and more expensive. Highway robbery is a better payoff than burglary as you get the whole truckload and the truck to move it, so you also have to pay for protection. That might mean protection as in guards, or protection as in racket, and quite often the distinction is far from clear.
This also becomes a typical first world GSM operations engineer's problem as soon as a big storm knocks over a few hundred towers and outs the electricity, as some bright spark inevitably notices the backup generator running.
Although you can buy solar and wind-powered base stations, there are still a lot of diesel ones out there. Now, if your objection is not merely financial, this means it's easy to destroy the infrastructure - you force open the valves and set it on fire. Interestingly, though, the Taliban have moved on from just starting a fire to breaking into the equipment cabinet and soaking it with the fuel, then setting that on fire. Thus multiplying the cost of repair and the downtime by an order of magnitude at least.
Alternatively, they sometimes dig a hole and blow the whole thing up with high explosive, wrecking the civil works (budget for quite a bit more including the labour) and demonstrating their aggression to everyone in earshot.
It also looks like they've realised that the backhaul links from the base stations to the switching centre are point-to-point microwave ones, and that the network has a hierarchical structure, with multiple base stations linked by microwave radio to a base station controller (or radio network controller in 3G) site which has a microwave link to the switch, and where there may be a variety of other equipment depending on exactly how the network is designed. As all that suggests, this is a crucial node and therefore a target. It is suspected that they have expert advice.
So the operators shut down service, and then the Afghan government and NATO yell at them to turn it back on.
And this is where it gets interesting. NATO has been installing macro-cells - big high power base stations - on its outposts as well as the private, ruggedised femtocells I wrote about with regard to Mr. Werritty. The idea was that if the commercial network was down, the phones would roam onto the backup network. Take that, forces of Islamofascism! But there's a problem. The commercial operators won't let the new network be in the list of permitted roaming networks on their SIMs, because they fear that if they shut down and service is still available, the Taliban will blow up even more of their stuff and perhaps start murdering engineers.
The government network could run like an IMSI catcher, masquerading as all four networks to capture their subscribers but forwarding everything - but I get the impression the operators don't want to interconnect with it, so calls would have to be routed out of the country and back in via the international gateway and it probably won't work very well.
And as a result, NATO has created the exact opposite of a successful emerging market GSM operator. Rather than cut-down low-power small cells cunningly distributed in the landscape, it's got big expensive pigeon fryers placed whereever seems safe or rather less unsafe. You'd think the same sort of place would do for a radio station as would do for a fort, but radioplanning is far more complicated than just picking hilltops and often deeply counter-intuitive. Rather than rock-bottom cost-to-serve, it's thought to be the most expensive phone network in the world per-user.
It's possible, thinking back to Rory Stewart, that a network designed along the lines of the kind of wireless-mesh broadband system his mates are building for the Penrith area might be more robust against such an attack. The Mexican Zetas seem to think so. Even staying in GSM, the BSC functions can be forward-deployed to the cell sites, and more of the backhaul could be point-to-multipoint rather than point-to-point, and more of the sites could be interlinked, thus getting more redundancy at the expense of worse efficiency. But that would only reduce the number of critical nodes. GSM remains a fundamentally hierarchical network architecture, and some would inevitably be much more important at the system level than others.
And finally, they could still just destroy towers, only with rather less efficiency. Putting more equipment at the cell site might just make it more vulnerable. Also, a problem with mesh networks is that they are more effective the more nodes there are - but the places where we usually want them because other networks are impossible tend to be sparsely populated. It would also make the whole issue personal. Owning the device would make you a target.
In the final analysis, fire remains an effective technology of rebellion.