Rumspringa
In the last few weeks I’ve been feeling an odd nostalgia for the paranoia of my past (a paranostalgia?). As if we’d actually slipped into a future of non-stop bad craziness I’d once darkly projected. The answers to a decade’s worth of “so what happens when…” started arriving, one at a time. And just didn’t stop.
Really, I ought to feel better prepared.
Way back, before that point in the mid-90s when “internet” became a thing that you had in your home, my networks consisted of dialing-in, off-peak, to Bulletin Boards. Furtively, and with haste. Phone calls were charged by the second and modems were slow. Every file came to represent an investment of time and money (metered calls, puny baudrates). And for the most part the information was of an underground nature, I mean… why bother downloading information you could read in a magazine or an easily findable book? HaXxor philes, all manner of New World Order conspiracy missives, drugs and magicks.
A small box of floppy disks filled with samizdat. Replicated. Traded in the schoolyard sneakernet, peer to peer. One day, “they”, would find out what was going on and they’d take away our modems. And while that might sound quaintly naive with distance, back in 1994 many of the Italian BBSes were seized by armed police during the night (shortly after Berlusconi had come to power). Even without the net we might distribute them on disks, keeping them alive like the living books from Fahrenheit 451, so that the people of the future would know about… smoking banana peels or whatever.
Then a zmodem’d copy of KA9Q. The actual Internet. Like watching the introduction of Australia’s first rabbits.
Crypto was one of the early worries. In the early nineties anything above 40bit crypto was considered munitions by the US government. When the PGP code was distributed via Usenet, it’s author Philip Zimmerman was considered akin to an arms smuggler and investigated and intimidated for several years, but not charged. It was released in response to US government plans to ensure there was a government back-door into all available encryption systems, something that European governments (UK especially) were also pushing for in the mid 90s. That possible future hasn’t yet come to pass. And while almost everyone employs some level of encryption every day they use the net, I still feel a little ashamed for using my PGP key so infrequently that I eventually forgot its pass-phrase. Most messages sit there, in plaintext, in the cloud. A secret subpoena away from being revealed. Or a snarfed password.
We’d worry about the state security agencies using Echelon to read our digital traffic and automatically build profiles, but by 2009 we discovered private companies were doing the deep-packet inspection (and injection) to build advertising profiles. Not too far removed from the recent activities of former Tunisian government, using DPI and injections to steal their citizen’s passwords and remove online advocacy resources. More often than not, those resources existing primarily as a means of… compiling and selling profiles to advertisers.
And while some spook or hacker could always have been monitoring your traffic, by 2010 a simple browser plugin made it reasonable to assume that anyone in wifi-range was doing it. We’ve just been riding the arc from “science fiction”, to “like something from science fiction”, to the everyday mundane. If, ten or fifteen years ago, you’d have suggested it wouldn’t surprise me to discover my mother’s PC had been hijacked by Russian botnet gangs, and used to hawk pharmaceuticals, I’d probably have found the idea of my mother owning a PC alone as implausible – the rest strictly-Gisbsonesque. (When evoking William Gibson implied the future rather than the observed present, of course.)
And then Egypt, which we’re all still unpacking. If nothing else, it’ll have enshrined the “Internet Kill-Switch” as bogeyman for generations to come. The US governments reaction to WikiLeaks, at the very least, removed a lot of the “it couldn’t happen here” assumptions that the internet was somehow beyond powerful influence.
It’s a shameful confession: governments messing with communication networks are more likely to get my attention than if they engage in a sustained campaign of violence, intimidation, and torture. I’m not sure what this means. Possibly, I’m a monster.
Prior to, well, last week my assumption would have been that sites and fixed locations on the net could be vulnerable, perhaps certain narrowly used protocols might be targeted. But the entire network? The counter-argument to any theoretical state-ordered net shutdown in the US is obvious, the economic damage would be considerable. No country of a large enough size and aspiration could possibly risk such action. Messing with the net is messing with the money.
That’s why it’s always felt stronger than whatever anarcho-pirate-darknet we might have ended up with.
Yet Douglas Rushkoff has been sounding disappointed, and somewhat exasperated, with how the internet’s turned out.
“From its Domain Name Servers to its IP addresses […] its flawed, centralized architecture made it ripe for conquest.” Its central brain apparently infected by a corporate parasite, some Toxoplasma Negotii, compelling it obey only the routing protocol of Mammon.
I’d strongly disagree with this characterisation “highly centralised”. The centralisation of the internet is represented (by ICANN/IANA) in agreed protocol standards and registries of delegated authorities (which are themselves further delegated). There’s no centralised mechanism that dictates the path a packet travels across a building or across the world. It’s a network of interconnected networks.
There’s no single authority that claims authority to molest both a domain name under .cn and one under .ch, they’re separate authorities in different jurisdictions. (And although gTLDs such as .org and .com have been assumed to operate under US legal jurisdiction for decades now, not a month goes by without some foreign entity expressing shock that this could be the case.)
That consolidation and market forces have placed much of the critical application infrastructure (Amazon, Google, et al) in the hands of companies either in the jurisdiction of (or at the economic mercy of) particular governments only reflects weakness (or, arguably, rationality) of the users, than the network itself.
Rushkoff goes as far as proposing we “abandon the Internet” and set up some kind of 21C Fidonet. Presumably one in which the transmission method is free from government or corporate influence. So not via the phone lines then. Or pretty much the entire electromagnetic spectrum (aside from those tolerated unlicensed but regulated, frequencies). I eagerly await Mr Ruskoff’s future suggestions in this matter. Perhaps by somehow transmitting QWK files over the “morphogenetic field”?
(My theory is, on a long enough timeline, all writers covering tech eventually become Clifford Stoll.)
Which is not to say I don’t support moving internet tools toward more decentralised networks covered by more appropriate controls and jurisdictions. There’s been plenty of fresh activity in that direction in the weeks since WikiLeaks got yanked off the net (first by a DoS attack and then by the remarkable power of suggestion exercised by US politicians). Activists deploying modified routers and hacking on ad-hoc mesh routing algorithms. Slap an 8-bit-pixel Jolly Roger sticker on it and secrete it in some urban loft.
I heart that stuff, really. It triggers an endorphin release from my residual cyberpunk receptors.
But I’ll probably never use them myself, not “in anger”. Networks are strengthened by network effects, so even tiny barriers to joining (compile this daemon, flash this device, etc) weaken potential. You end up needing recognised standards groups (or market dominance) to encourage adoption. And activism or rugged grid-independence are not quite compelling use-cases for corporate product development.
It’s easy to recognise the point in which a technology reaches the point of mainstream acceptability, when it’s “good enough”. It’s the exact same point at which the geeks lose interest and move on to something else.
Most mesh networks in the future won’t be initiated by tinkering with laptop settings… it’ll be someone powering up a device in the office that’ll latch on whatever the nearest station is. A printer perhaps, a random piece of Cisco, an HP thingum. Likely all talking IEEE 802.11s with whatever the lowest-common-denominator routing is (HWMP).
You’ll barely notice the OS update. The extra tab on your router’s web config. Smart-phone vendors unworried by battery-life consumer experience will add another unread bullet-point to their feature lists.
You’ll buy cheap little things you plug into power outlets with no (or almost no) configuration or administration. Just wi-fi network extenders pitching the use of the home office router connection with “the TV in the den”. Unseen black boxes, devoid of interface, devoid of emotional connection. A rapid path to becoming smaller, more efficient, cheaper… almost to the point of disposability.
Then one day adverts for garden wifi extenders are routinely seen in the back of the Sunday magazine supplement. Cheap solar-powered mesh stations. Foot-long spikes topped with teacup-saucer photovoltaic cell, like those solar garden lights. Picture network Johnny Appleseeds planting these things as they move around, geo-tagging as they go.
But starting with the network and waiting for the apps to follow seems like the wrong way to go about things. Post-Napster filesharing is already way ahead here. In the conventional realm, ad-hoc tools exist for Zeroconf discovery (Bonjour chat in iChat, collaborative editing in SubEthaEdit). Does that scale to something like a completely decentralised ad-hoc equivalent to Wikipedia? What does an ad-hoc Facebook look like? Plexus? Distributed timely local information and mapping for example. Walking around with an iPhone, you may be 50m away from something, but unless you’ve got a strong data connection available, you can’t access the mapping information from some internationally-hosted server to check.
One of the main problems everyone’ll hit when attempting to marry an ad-hoc dynamic network structure with the Internet architecture is not necessarily with the hierarchical distribution of IP addresses of domain names, but the design decision that meant that IP addresses were used as both host identifiers and network locators. Not a bad decision in the 70s/80s where nodes were fixed points, but became increasingly problematic the moment the likes of DHCP, NAT and wi-fi were thrown in the mix.
A simple use-case might be safely corporate. An executive takes a critical video call (over IP) on her hand-held device as she’s leaving an office to catch a train. She continues the call in the cab to the station, then in the station, and during the high-speed train journey to another country – at each appropriate point switching over to whatever IP network is available or most suitable. Battery and bandwidth issues aside, this wouldn’t be possible under the conventional IP model. Even though the hardware and user remain the same, the network-centric identity changes for each different network which means disconnecting and reconnecting each time – with attendant authentication issues. Imagine a mobile phone system that needed to hang-up the call and redial every time you wandered between cells.
Letting the application layer keep track of network layer identifiers is a little broken. But for the last few years engineers have been working on separating them. The most promising solution seems to be HIP, the Host Identity Protocol. (Not least, because it’s got adorable acronyms such as HICCUPS and HIP BONE.) I’ve no expert knowledge, but this is what I’ve understood from skimming several RFCs.
Off the bat, the main downside is clear: it requires modifying the network stack of every end-point system that wants to use it (changes should be mostly invisible to the network infrastructure though) and any application for which network identifiers are specifically needed rather than host identifiers (e.g. networking tools). Not a problem if you’ve got mature code, but unsupported systems would be out of the picture. It’s not all-or-nothing though, if you need to connect to an unHIP host it’ll fall-back to the direct IP method.
HIP works by establishing a new layer in the stack between the Transport and IP layers. The HIP host creates its own identifier by generating a cryptographic key and identifying . If fact the host can have multiple identifiers and they can be as permanent or temporary as desired. Because they’re mathematically created, rather than allocated from an administered pool, they’re effectively disposable. A fingerprint is also generated from the identity, called a HIT, that’s cunningly formatted to look like an IPv6 address located in an unroutable address range.
So when a HIP-enabled host (mobile.example.org) wants to contact, say, a HIP-enabled server (static.example.com) the following takes place: the site attempts to resolve static.example.com to a HIP identity, e.g. by looking it up in DNS but checking for a “HIP” record type (falling back to the usual “AAAA” or “A” records for unHIP sites). If an IP address isn’t available in DNS, the HIP record can be used to discover the IP addresses of the host’s “rendezvous” servers which are basically the HIP-equivalent of a DNS server, keeping communicators updated about which IP addresses to use. The client then makes a HIP connection request to the server that involves them cryptographically authenticating each other. In the application layer of the server, the originator of the request is not a real IP address, but rather the IPv6-like HIT (and the same on the client). Any messages sent to a HIT address are caught at the HIP layer and specially encapsulated.
All seems a little complex to keep a long mobile video-call connected, but there are considerable benefits to widespread adoption. Firstly, multi-homing – the same host identity reachable via multiple networks, something that’s only usually achievable at the ISP level. Right now it’s possible to maintain multiple network addresses on the same machine, but they’re essentially separate identities. But imagine a multiple device that could slide between preferring a fixed wifi address or 3G based on a drop in signal strength.
Also, simple access controls. Given both sides need to authenticate each other, it’s possible to restrict access to specific HIT addresses since they’ll be the same regardless of which network is being connected from. Essentially, it’s an ad-hoc VPN. And, given both hosts HIP are mutually authenticated, there’s a secure channel for exchanging a session encryption key. So, by default, all communication is encrypted at the HIP layer rather than requiring all application-level protocols be amended. Also, the handshake mechanism helps defend against certain classes of DoS attacks since forged packets can be identified and discarded more easily.
And, for everyone still claiming that peak-IPv4 is “just a theory”, there’s built-in NAT-traversal to allow servers to run from private ranges.
Mobile clients making use of rendezvous servers on relatively stable and reachable addresses mean that they still benefit from relatively fixed chunks of infrastructure to operate. And also there’s still a place for DNS and the domain name allocation infrastructure (and their inherent problems) while the use of human understandable unique identifiers are considered useful.
However, since host IDs are (potentially) permanent, there shouldn’t be a problem with aggressively caching them at the client end. In fact, the host identities themselves represent a decentralised, distributed, naming scheme that doesn’t require DNS to function. For example, via distributed hash tables.
Fully crypto’d, allowing graceful transition between a variety of networks, based on decentralised and distributed addressing schemes. Isn’t that beginning to sound like the foundations of that new and better Internet we’d want to run to?
