FTP isn't very sexy any more. After a long reign, it is overshadowed by its step child, HTTP. Web page delivery (and all those status codes and headers) paradoxically turned out to be a superset of uncomplicated general-purpose file transfer. FTP is even slowly losing ground on it's home turf of getting large bags of bits from A to B with the least possible commotion. Peer-to-peer (P2P) file transfer such as BitTorrent is a more attractive way of distributing large binary files without running up bandwidth bills or being bottlenecked. Grandchildren like SIP have fled the nest and don't visit as often as they used to.
But FTP will go down in the history books as a technology that changed the world. Why will is not be forgotten? Because it filled a need complementary to another phenomenon of its time, open source software (OSS). I can remember being a coder out on a client site as recently as 1994 and gagging for a copy of emacs. I didn't have any questions as to what text editor I needed since the alternatives were too awful to contemplate.
In the end I managed to grab a copy from a cover CD of a magazine, and carried that CD with me for several years afterwards -- a decent scriptable text editor saved the day on several occasions. I'd been using the Internet and FTP since 1989, but such connectivity was unheard of outside of an academic environment. GNU software was an obscure backwater of Unix systems development, and Linux had barely crawled out of the primodial bit soup. Nobody expected to get an Internet connection to be able to do software development.
FTP (as will as HTTP, P2P and SIP) would not have become ubiquitous if we were made to wait for a telco to build a file transfer network for us. We'd be waiting for the marketing department to to focus groups, market segmentation and pricing plans for every file type. The network group would have gold-plated the implementation (at ten times the cost) to include guaranteed delivery for high-end business clients. The standards groups would have won a jackpot on their frequent flyer miles to Geneva and Hawaii.
The value chain of OSS wasn't complete without two additional essential elements. The first was cheap networked distribution unencumbered by telco service-based price discrimination. Indeed, OSS is a two-way street -- someone has to write all that software as well as download and install it. The supply side was bottlenecked by access of the general coding public to technical resources, and to development community co-ordination sites like Sourceforce. If you were around in the era of bulletin boards, you'll have a vivid memory of what sort of cost and difficulty communicating in the pre-Internet era was like.
The second part of the value chain was distribution. For this you needed a pervasive protocol for file transfer, as well as universally available and interoperable server and client software. Generous universities (often funded in the background by hardware vendors like Sun) provided hosting and connectivity. FTP was (and still is) the technology that fills the distribution need.
The explosion in popularity of OSS has seen most of the software you need to perform everyday computing tasks fall to a price of zero. The exceptions are at the high-end: big iron databases, critical enterprise operational systems, and the unglamorous middleware that glues it all together. Even Microsoft Windows, despite the billions of dollars of cash it throws off, doesn't cost nearly as much as you would expect from a product in such a monopoly position. The wicked monopolist can't constrict supply to raise prices because there is an unlimited supply on hand of a good-enough substitute OSS product.
In short, the stupid network concept enabled FTP, and together they enabled OSS to occur and change the world of computing.
But the increase in supply and usage of OSS has seen its own problems. And we can probably learn from this experience, since the problems OSS products have had in a world of zero marginal cost and superabundance may come to afflict communications services in the same way. The same dynamics are in place, even if the suppliers offer proprietary solutions.
To take a product example, I recently decided to install an OSS spam filter on my home mail server. There are plenty to choose from. In the end I picked DSPAM (nice software, shame about the documentation). But the process of getting the right thing for my needs was frustrating and time-consuming. The problem is that at a price of zero the concept of marketing disappears. OSS has torn away the marketing barriers, and sites like Freshmeat and Download.com haven't really filled their place. Who is going to pay to analyze the needs of people like me, and pay to promote their message ahead of every other marketer who wants my attention to switch to their message?
Sites like Freshmeat do make an effort. They have download popularity indexes for different software. There are feedback scores. But there's a huge amount of noise from abandoned, irrelevant, duplicative or inadequate OSS projects. And ultimately, they aren't attempting to have any form of relationship with me. There's nothing that says you've already downloaded Postfix, Apache and Courier, and run a a two-user mail server accessed from multiple clients, so maybe server-based DSPAM is the spam filter for you.
The fully commercial Download.com server lacks context from which to advertise to me. All they know is I'm looking for some anti-spam software ('cos thats what I searched for). I'm not in the market for shareware (those Scottish genes keep my wallet tightly shut). The one thing they can't advertise is anti-spam software! So they're left with generic stuff like cars and gambling adverts that don't generate enough revenue to do any in-depth relationship marketing.
Lastly there's the supplier of the Linux distribution I happen to use (which happens to be Mandrake, selected largely at random). In principle they should be the ones bundling and marketing a pre-integrated OS to me. But their marketing department seems to have omitted to address the home server market. And one glance at their web site will tell you that marketing communications isn't their strongest point. Plus, of course, they've utterly failed to get their product pre-loaded onto PCs and into mass-market distribution channels. Blaming Microsoft sounds like a lame excuse for inaction, and a lack of boldness and innovation at distribution. There are plenty of alternative white box manufacturers and retails channels. Just ask Lindows and Wal-Mart.
At the end of the day, most core communications services boil down to a list of user authentication and profile data, a bunch of session protocols (remember our friend, FTP?) and a graphical user interface. If your data isn't particularly proprietary, and your protocol and user interface aren't plastered with patents, the service can be replicated easily. And the experience of Skype suggests that some (if not yet all) of the service can be distributed, removing all operational cost. Although Skype doesn't distribute authentication and routing, these will happen too with time. We're already seeing things like distributed P2P disk backup that were unthinkable until recently. Trusted computing technologies might have a bright side and resolve a lot of the integrity and security issues of distributed computing in an unsafe world.
Once you've eliminated the need for a central server and associated costs, any kid with a text editor and compiler can become a phone company. Communications adopt the same economics as open source software. There might be some rough edges, but these services will be good enough for most friends and family use or commercial internal use. They'll certainly fall within the acceptable performance range for a free product. They'll get better over time and subsume paid-for products.
Then Voompf! An explosion of simulacra communications services. None of which is popular enough to support the secondary ecosystem of services and customization that drives adoption success. We've seen this (writ small) with the X Window system and the infinitude of Linux desktops. Or photo albums. Or email clients. Or text editors. Ad nauseam.
Some will have proprietary data or technology and won't be OSS'd. Some will be too nimble for the OSS crowd to keep up. Some will tie their offering to some other money-making enterprise (which could be pipe provision, or server sales). Niches will need to be filled. There's enough of that to make Eli Noam's predictions of infopocalypse unlikely to be fulfilled. The low-end of the market may disappear, but new layers will always be added on top. But basic communications service (*not* connectivity) when implemented in a distributed peer-to-peer fashion is fundamentally a non-rivalrous resource which is to be priced at zero.
The advantage of the PSTN is that there's only one to choose from. Everyone orbits the same target. The fatal flaw is the same. Nobody can orbit a new target. If an endless cacophany of "free" communications services (VoIP, push-to-talk, IM, chat, etc.) continue to proliferate, the opposite may occur, and users may simply throw up their hands in frustration. They'll stop thinking for themselves, and submit their communications needs to an aggregator like AOL, Yahoo! or MSN. These are guaranteed to have other agendas than maximized user communication freedom and service innovation. (Think: futile media cross-selling, adverts and Windows lock-in respectively.) We're already seeing multiple Skype-alikes in the offing.
The biggest problems of OSS come from the weak execution of the functions usually done by sales, marketing and distribution. Understanding user needs. Communicating the existence of solutions to those needs back to users. Getting the solutions delivered into the user's hands. That isn't solved by any fancy new protocols or technology, since at heart it's a people coordination problem. If you are a young coder today looking to make a name in OSS, please bear in mind the biggest unsolved problem today isn't yet another son-of-a-son-of-FTP.
The same problem may be about to afflict personal communications. For the communications revolution to meet its true potential, it's time to think of how to automate ("open source") marketing. All those cell phone stores and TV adverts aren't for nothing. Believe it or not, that distribution system has a useful purpose. Only products that have real value get to cut through the clutter and justify their marketing and distribution costs. But once you cut out the telco rent seeking, how does the customer get the service they need, even if it is free?
One scenario is that the market just works its magic. Apple comes along with a seamless package of iChat etc., you pay your dollar to Mr Jobs and come away happy. Or someone like Google launches a better service, the message spreads fast by word of mouth without any marketing effort, and they win the race. Or Lindows innovates the distribution model and (maybe) cracks the market.
On the other hand, you end up with a dozen competing Skypes, just like you have a dozen competing Linux distributions. You end up in a dead-end coordination failure. Low barriers to entry, abundance and competition are good only as long as the competition leads to clear winners and losers.
To summarize, there's a danger that IP communications will veer down the path of open source software and become too fragmented. This will be because of a market breakdown in the absence of paid service. User demand cannot be communication to suppliers through the price mechanism, and suppliers that fail to meet user demand are not easily eliminated. The result is an unintended consequence of reinforcing the power of paid-for services in adjacent markets (e.g. hardware, operating systems, content and connectivity). This fate is far from inevitable, but the threat is real.
Posted by Martin Geddes at 5:37 PMTrackBack URL for this entry:
http://www.telepocalypse.net/cgi-sys/cgiwrap/mgeddes/MT/mt-tb.cgi/189