Having spent too much time in airports and on planes again recently, I've been pondering the open-vs.-closed issue for networks.
We often get confused by the words we use, or the lack of them. For example, is the word "Skype" denoting a company, a product, a verb, a network, a protocol, or all of the above? Similar considerations led me earlier to coin the word "Vob" to describe the network of provisionless open SIP-speaking telephony nodes in distinction from the SIP protocol itself. (For instance, just because it's SIP doesn't mean it's telephony or the Vob. SIP is used for other stuff besides the Vob.)
Sometimes the network is merely an emergent phenomenon, such as the Internet. It has been argued that the term "Internet" is really a synonym for the transitive web of social agreements to transit IP packets, as opposed to any concrete thing. We could change the protocol (and with IPv6, we would) whilst retaining the social agreement, and it's still the Internet. (Thus the definig attribute of "Internetness" can't be determined by solely technical observations -- you have to look at the social and economic agreement to see if something is really "on the Internet". If your ISP's terms prevent you from acting as a transit, you're only an associate member.)
Other protocols, such as SMTP, give rise to similar confusion. Are you really "doing email" when you send a message inside your corporation using a protocol other than SMTP? And what if you aren't able to relay that message to all the rest of the world? We don't have a word to describe the "SMTPNet", so people don't really notice when they're only being given partial access to it.
Is there an open network of blogs (such as this one) able to accept trackback pings? The Trackbackosphere?
Ted Shelton spots the idea of a storage-centric network. There is an implicit network of devices that are capable of transmitting and playing tracks purchased through iTunes. This is a closed network. We don't have a name for it, although we intuitively understand it.
I recently saw similar confusion over IMS. For the uninitiated, this is the carriers getting together to plan a universe of smart network elements upon which you will be induced to become hopelessly dependent and pay eternally with your soul and wallet. But IMS is just a technology; what's the network? Will there be one IMSNET? Kindasortof, since the spec talks about foreign agents and roaming. But won't the carriers want to differentiate their service offerings? So will there be lots of little IMS-speaking networks with local and mutually incomprehensible dialects? At least we know whatever the outcome, it'll be closed.
(As an aside, the option value of an IMS network seems very low, whereas an IP network is practically unbounded except by capacity. Lord knows what the carriers plan to do to make this vision compelling -- or compulsory -- to their customers.)
BitTorrent is a protocol, although you often see people get confused and think it's a network. There are as many BitTorrent networks as their are tracker sites. Most are "semi-open" -- anyone can become a client, although the content of the server may be restricted. I'm sure there are plenty of darknet torrent nodes too, which are completely closed. Napster was very (too!) open; KaZaA is pretty open still.
So, what determines whether we get an open or closed network architecture? I'm not sure, and don't have any research, but I do have some hypotheses. I suspect that there is no simple answer, but rather a basket of competing forces.
There are (at least) two forces pushing towards a closed architecture.
Firstly there is the need to preserve the integrity of the network. For Apple, "integrity" means copyright control. For Skype it means controlling telephone spam. For KaZaA it means battling poisoned content distributed by the RIAA and MPAA. The deciding factor seems to be a "before/after" switch: is the damage done as soon as the "bad" message is received, or can the badness be filtered out later? If I can copy the iTunes file, the damage is done; too late to verify if it has remained in (RIAA-)safe hands. The interruption can't be undone. The poisoned file can maybe be filtered out later, hence the permeable architecture of most file-sharing sites. E-mail can use after-the-event filtering based on message content; telephony cannot.
The second factor is creating a consistency of experience. If you control all the nodes, you can assure senders of the experience that receivers will enjoy. In Skype, for example, you know how the ringing experience will unfold, what sort of "caller ID" they'll see, and what filtering rules everyone has access to. That didn't exist on the Vob. [After I wrote this, but before I posted, I saw a similar sentiment on Jeff Pulver's blog. But fixing compatibility and consistency doesn't automatically give the Vob parity with Skype in the user's eyes, IMNVHO.]
Pushing towards openness is the option value of the network -- its "stupidity". SIP is a kinda crazy hybrid. It's a flexible technology. But it's hardly what you would call a "stupid network" approach. Just take a look at the abstract from the RFC.
SIP makes use of elements called proxy servers to help route requests to the user's current location, authenticate and authorize users for services, implement provider call-routing policies, and provide features to users. SIP also provides a registration function that allows users to upload their current locations for use by proxy servers.
Lots of tightly coupled functions in cental nodes, with some stateful sauce sprinkled on top for added flavour. (Has anyone else noticed that you can get 50% of the benefit of SIP for 2% of the cost using dynamic DNS?) Whilst the remaining flexibility has had an unfortunate effect of creating chaos for users, it still makes it quite a useful toolkit to plunder for building your closed network. The temptation to fix a few of the wrinkles and holes will result in a proprietary dialect each time.
Skype's basic framework is only as flexible as Skype's internal coding team. But the API enables localised customisation without compromising the integrity of the core network. SMTP enabled the emergence of hundreds of types of clients and dozens of mail servers; if it hadn't been so flexible and open, it wouldn't have succeeded.
Completing the picture is the level of market power the owner of a closed network has over its users; the amount of tribute demanded as a consequence; and the availability of viable alternatives.
I'm pretty sure I've not captured the whole picture here. I'd be interested to hear reader feedback on what drives networks to be open or closed, or references to academic study on the matter.
Hmmm ... I wonder what's the future of openness and mobile operators? I rather liked this quote on the matter, if in a somewhat restricted context:
I believe in the end open distribution models will win. In a way mobile operators are like Michael Gorbachev in the Soviet Union in 1988: they are already preparing to promote perestroika in an effort to control the explosive market forces. Eventually even that will fail because the consumers will vote with their wallets. With some luck we might even see the iTunes phone one day, 'Carrier willing'.
Ah, the irony of it. The carriers will be forced to open so we can enjoy our closed iTunes and Skype networks.
Posted by Martin Geddes at 11:33 PMTrackBack URL for this entry:
http://www.telepocalypse.net/cgi-sys/cgiwrap/mgeddes/MT/mt-tb.cgi/461