Interesting People mailing list archives
Purpose vs discovery and the Internet as a dynamic
From: Dave Farber <dave () farber net>
Date: Tue, 3 Nov 2009 12:36:18 -0500
Begin forwarded message:
From: Bob Frankston <Bob19-0501 () bobf frankston com> Date: November 3, 2009 12:33:11 PM EST To: "Prof. David J. J Farber" <dave () farber net>, nnsquad () nnsquad org Subject: Purpose vs discovery and the Internet as a dynamic
I’m writing this in response to the myriad discussions about how to make sure that the Internet continues to “work” despite P2P or whatever the current threat seems to be. Behind much of the discussi on is the presumption that the Internet has a purpose in the sense o f making some applications like video games and VoIP work. Yesterday we feared modems, today we fear P2P.This confusion arises from the success of the Internet. People see what works and assume that the purpose of the Internet is to support those applications. The fact applications that require low latency (delays) and low jitter (variability) work in some cases means that we must now promise to make sure they work everywhere. The very innovation that made the application work so well at such a low cost is now seen as a threat to the new status quo.We “prove” this by example – but that’s just the opposite of science. Science is about testing ideas rather than looking for conf irmation of our presumptions. We then add more confusion by building in the mechanisms we presume are necessary thus which, of course, “ proves” that they are the reason the applications work.In fact the applications started to “just work” because the mechanisms are not built in – it’s this seeming paradox that keeps us arguing in circles. As I explain in http://frankston.com/?n=InternetDynamic VoIP did not work in the 1980’s. Or, to be more precise, it might’ve indeed worked over local networks but you couldn’t presume that it would work between LANs, especially when dialup conn ections were involved. Instead we used the Internet for more toleran t applications such as email and file transfers.If we had to make voice work we could’ve built it into the network b y being dependent upon the network giving us a dedicated path. This is the basic design choice made in SS7. Both the Internet and SS7 we re done by CS people but with different assumptions. SS7 achieved it s goal and supporting high quality (56KB in the US) voice but at the price of being dependent upon high priced gear (as per 1970’s price s).By eschewing dependence on such gear the Internet (as a thing) could n’t make such promises. Instead we had to find out what worked and g o with it. Email worked because it was very tolerant and typical mes sages were a few hundred characters with people flagging messages ov er a thousand words or so as large.Experience suggested that applications such as voice and video could n’t work without special help because they were too sensitive to lat ency and jitter.Then the Web happened. The web itself was initially about text so the latency and jitter weren’t issues but it did generate a lot of t raffic. And people predicted the Internet would collapse. Instead th e opposite happened – demand created supply. One reason is that we’re able to take advantage of any available bits with congestion b eing an annoying but not fatal. For example lost packet used to clos e a click but we’ve learned how to smooth over such problems.It turned out that the capacity increased quickly where there were no disincentives. The increase came in many forms which can be loosely called Moore’s law effects. But as I wrote in http://rmf.vc/?n= BL the physics aspect of Moore’s law is secondary to my marketplace formulation. If you decouple systems and embrace any opportunity you get hypergrowth. It didn’t matter if you made bits run fast or you had more paths the net effect was more. (no puns intended but you ca n find them if you wish).Thus fiber in the ground was able to carry more bits thanks to improved gear at the end points. Yet DSL didn’t improve much beyond the initial 1980’s implementation of ADSL because the carriers had n o incentive beyond the original purpose of Interactive TV and improv ements threatened their ability to charge high prices for T1 lines a nd for bits themselves. The same thing happened with fiber when the improvements led to a glut in capacity about 2000 and the carriers r eacted by purposefully limiting capacity and then pretending the lim its were inherent.Meeting the demand for “web bits” had a side-effect of giving us copious capacity and the low latency, low jitter application started to “just work”. But we mustn’t forget that this is a process of discovery not one of promise. We might argue against the asymmetr ic “broadband” connections but we also discovered that they are indeed useful for video, especially since that was the original purp ose of the underlying architecture.To make this even more confusing is that once VoIP started working over high performance links clever folks started making it work seeming unsuitable paths and then voice-grams started to blur the distinctions between conversations and messaging. On the surface it all looks like telephony but it’s not – and trying to preserve telephony becomes counter-productive.The problem today is that observers who see these applications “just work” confuse discovery with purpose and want to bake in the applic ations in the same sense that some people supposedly wanted to close the patent office in the 1800’s because everything had already been discovered. (A useful story even if not true).I do think that we need new protocols less dependent upon network managers tilting the playing field and protocols which limit the ability of “bad players” to prevent others from discovering new possibilities. This is one reason I’m wary of “proper network management” since “proper” can reflect a presumption of purpose.The fears of network collapse have a basis in reality just as the warnings that modems would destroy the phone network were real modulo their assumed architectural limits. But the solution was not in better network management and was instead in increasing the capacity so such problems became moot.As with my comments on DSL the issue of incentives and funding are fundamental. Today’s network in which network operators are threaten ed by abundance is the new “modem crisis” with concerns about neutrality being a countervailing force.The real danger in the purpose-driven network is that we focus on how to manage scarcity by favoring applications on the presumption that the “network” is making promises rather than providing opportunity. We should be addressing the root (and route?) causes of the problem – the very idea that carriers have to monetize the traf fic and thus have every incentive to limit capacity and make promise s they that they can charge for. Indeed carriers proving bandwidth a nd video are making promises rather than just providing opportunity.Alas the discussion of “neutrality” gets lost in this controversy over purpose. I’d like a simple formulation in which providers have no stake in making promises but the discussion may be too polluted b y the presumption of purpose.These considerations are very much on my mind as I write about Ambient Connectivity. It’s about creating opportunity and decoupling the applications from providers who need to monetize each path.Understanding AC requires a nuanced understanding of the Internet’s dynamic and the success of the experiment in decoupling the applicat ions from the accidental properties of the transport and its “owners ”.
------------------------------------------- Archives: https://www.listbox.com/member/archive/247/=now RSS Feed: https://www.listbox.com/member/archive/rss/247/ Powered by Listbox: http://www.listbox.com
Current thread:
- Purpose vs discovery and the Internet as a dynamic Dave Farber (Nov 03)