Date: Wed, 12 Feb 2003 08:51:02 -0800 From: Terry Lambert <tlambert2@mindspring.com> To: Dave Hayes <dave@jetcafe.org> Cc: freebsd-chat@FreeBSD.ORG Subject: Re: Modelling complexity (was: Re: matthew dillon) Message-ID: <3E4A7B76.2D6856C6@mindspring.com> References: <200302121149.h1CBnp143597@hokkshideh2.jetcafe.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Dave Hayes wrote: > Terry Lambert <tlambert2@mindspring.com> writes: > > Dave Hayes wrote: > >> > I understood you perfectly. > >> > >> Then why did you paraphrase what I said incorrectly? > > > > I paraphrased the *consequences* of what you said, were less than > > 100% of us to adopt your world view. > > Technically, you paraphrased your assumptions about the consequences > of what you -think- I meant. Technically, I paraphrased the results of gross calculations about the consequences of several of the highest probability interpretations of what you meant, and then stated the common properties of all interpretations with "reasonably high" probabilities. 8^p. > Yes, this is your theory. You still haven't demonstrated how the > system is currently in a steady state, I have. You've just refused to accept/consult my references, and I don't have the time or inclination to teach you mathematics in this forum. > why any steady state is desirable, So that state transitions can be planned, of course. Otherwise, how do you know in advance whether a particular state transition will result in greater or lesser entropy? > nor have you seemed to grasp what I am advocating. This > could be because I am not explaining it in terms of chaos theory or > mathematical rigor. Or even in terms of the desired consequences of behaviour according to your model, or how such behaviours would be self maintaining, once established. Feel free to engage in a generalization, and permit me to decide on my own whether the steady-state you propose is actually desirable. > > You hypothesize that all behaviour is tolerable, and should be > > tolerated (if only by ignoring it). > > Not exactly. I maintain that the most efficient change to any > system you are participating in is the change in your behavior. > If you try to change other's behavior, it will not work for > some small select group of people. I maintain that the most efficient change to any system you are participating in is the change in the ground rules for the system, such that, given a system S, the resulting system S' has the emergent properties that you desire, which the system S did not. "Give me a lever, and a place to stand, and I shall move the world!" -- Archimedes > > Why don't you take your simple system, and posit a strange attractor > > of a willful disrupter, and see if your system maintains itself in > > some acceptable range +/- some small value, independent of the > > amount of force driving the occilations? > > Frankly, because people don't all work that way. They all *do* work that way, as a group; it's called a "mutual security game". And since what we are interested in here is the predictability of group dynamics, given a set of circumstances, such as an external threat, we can make useful adjustments to the system in question so that the group utilizing the system as a tool will be able to deal with the threat. The most absolutely crude form of this is "list moderation", but we are engineers, and, as engineers, our tools are better than a single on/off switch, which we must wield with an iron fist. > >> BTW, you cannot prove that you understood what I said and neither > >> can I. We might agree that you understand, but the possibility > >> would still exist that you do not. > > > > And how is that relevent? > > It's always relavent, any time someone is communicating. Failure to > account for this would be like failure to account for a key piece > of datum in your experimental verifications. Ignoring it's effect > is tantamount to religious behavior...for a real scientist. It's never relevent, in fact. Are you aware of NIM-like games? Ones such game is called "The Prisoner's Dilemma". Another version of this game is called "The Iterated Prisoner's Dilemma". The difference between the two is that in the iterated version, you cn take playing history into account. Now consider that the prisoner's in the game have no way of communicating, apart from knowledge of whether or not their counterpart has betrayed them, and the final score at the end of each iteration. What is the optimal playing stratey for this game? The answer is called "Modified Tit-For-Tat With Forgiveness" (you may look this up on Google with search phrase "iterated prisoner's dilemma"). The reason that this is the optimal playing strategy is that it uses the outcome of the game over a number of iterations in order to establish a covert communications channel, and thus to offer to cooperate to achieve the highest possible score. This communications channel was not designed into the game; it is an *emergent property* of the playing strategy, and it's what makes the playing stragey optimal. > > If you want to be a Solipsist, then you aren't really going to be > > useful to anyone else, because your going to act like everyone else > > is a figment of your imagination anyway. Where is *our* utility in > > *you* acting like that? > > Gee. If you -are- a figment of someone's imagination, would not the > greatest utility always be in interacting with the person who is > imagining you? ;) > > In fact, if I -was- imagining you, and I gave you such free will, > belligerence, and intelligence as you display, where would -my- > utility in that be? By your own standards, wouldn't I rather > imagine a bunch of people who agreed with everything I said? > > Point of fact is, I'm not really a solipsist. Being accused of this is > the method most use to convienently misunderstand what I am saying. Your argument was Solipsist in nature. In fact, if I were to ascribe a philosophy to you from my own opinion, it would be "contratian constructivist", where you insist everyone argue with you on the basis of first principles, but then refuse to give the same courtesy to the person you are arguing with, so as to try to make them "work for it". Thus the easiest path becomes agreeing with you, and getting you to accept any points which are not in correspondance with your worldview is a Sisyphean task. People must then weigh their attachment to their own philosophy vs. the amount of effort required to overcome yours; this approach is aided by the fact that, at any perceived mistake on their part, you force a return to first principles (disallowing return to a checkpoint), AND you will not go away until you brow-beat them into your point of view. Vis-a-vis, your insistance on picking up an argument that was ended over six months ago. That is, of course, just my opinion... ;^). > >> > By designing the situation, you design the constraints, and therefore > >> > you limit the possibilities for behaviour. > >> > >> Now you know why I don't take most of your analogies seriously. =) > > > > I have no idea. > > You design analogies to represent a concept you are trying > to substantiate. Unfortunately, you design the situation and the > constraints, which limit the possibilities for behavior to that > which support your concept. Yes, of course. To do otherwise would be to argue from the specific to the general, which is a logical fallacy. I argue only in terms of the limit of the problem space currently under consideration. It is an engineering approach to arguement: I do not like to solve a member of a class of problems more than once, if I can avoid it, so I will determine the boradest possible class which is capable of containing the problem statement, and I solve that, instead. Engineers will generally take this approach, if they are given free rein to solve a problem: it's precisely why an engineer will spend days writing and perfecting a shell script to perform some function which the engineer could have performed manually in less or equal time. By creating a tool suited to the problem, instead of merely immediately addressing the problem, when or if they or someone else are faced with the same problem in the future, they can treat it as already solved. If the have generalized sufficiently, then the same solution applies to a much larger class of problems than the original subset being solved for by the original engineer. The primary difference between a good engineer and a bad engineer is whether they solve the immediate problem, or if they solve the general problem, when given a problem to solve. When interviewing someone, I always pose at least one problem with an obvious general solution that is no less efficient than several "brute force" solutions. If the candidate picks one of the "brute force" solutions, then I know that they will be "Mr. Right Now", rather than "Mr. Right", and any code they write will have to be revisited later on, as the product, specifications, or user expectations for the product, evolve. I will hire this person if I need someone to create prototypes, and there is no better candidate, but not otherwise. Likewise, asking a question where the implementation cost is higher for the more general solution, if someone comes up with the more general solution, I will ask them for possible solutions with quicker times to implement. If they cannot give them, then I will hire this person if the specifications are written in stone, but not if I need an agile team member capable of rapidly creating prototypes. Some people can do both; even people you would pidgeon-hole as one type or the other, based on past experience, can surprise you, if you interview them this way. Generally speaking, in that case, your past experience is based on them being forced into one role or the other, because your team is otherwise unblanced. If you find this polarization happening, even with a number of people who can operate in both modes, then it's probably management that's at fault (e.g. setting unrealistic deadlines, goals, milestones, or specifying overy-large feature sets). Pointing the finger at management is always the last resort, in most cases, and as long as management feels confortable managing the people friction, they will often not do anything about the problem. But the organization, as a whole, will not be as efficient, effective, or have as high an economic return as it would otherwise. Time to recommend another book, for people who are recognizing their own organization in this discussion: Out of the Crisis W. Edwards Deming MIT Press ISBN: 0262541157 This book deals with organizations which engage in the practice of "crisis managment", rather than running smoothly and effectively. > > Is it just that you can't follow the mathematics behind emergent > > properties of chaotic systems, or is it just that you disagree that > > the chaotic systems can be constrained in such a way as to have a > > particular desirable set of emergent properties? > > Neither. I disagree that you can usefully model the world of > people's interactions mathematically. If you can do this, > why aren't you playing the stock market and getting so rich > that you can buy my agreement with your principles? ;) Modelling the stock market is not the same as modelling group behaviour. It's well known that you can model the stock market nearly perfectly. The formula is called the Black, Scholes, and Merton formula. Using this formula to price options, you can subtract risk out (or determine the level of acceptable risk, and the risk/reward ratio you are willing to accept t be paid for taking the risk). As for bribing you, I'm not so naive to believe I would be buying agreement, so much as I would be buying lack of public dissent. 8-). > > No. You are mistaking my use of the word "law" for the common use > > of the word "rule", as in "some arbitrary value whose compliance is > > enforced by a larger society". A "law" is "the way things are". For > > example, it is a "law" on the Internet that you can only send packets > > with certain protocol number values (IP, IPv6, ICMP, etc.). This is > > "the law", because no matter how hard you try to send packets with > > protocol numbers other than that, they will not get from point A to > > point B over an arbitrary segment of the Internet reliably, because > > the core routers will simply discard, not forward, them. That's > > "just the way things work": it is a law. Like gravity. > > So I hack into a router, upload a new core image, and now it transmits > what I want. Yet gravity isn't something I can hack into and change. > That difference was what I was attempting to describe. You would have to hack all routers intermediate to point A and B, and all routers for which there was an equal cost for intermediate hops, for all intermediate alternatives intermediate to A and B, and you would have to hack all routers that may be involved in failover, should you wish to maintain end-to-end communication in failure situations in which end-to-end communication would normally be maintained with unhacked routers. At which point, you might as well buy Cisco and make them do it, or take over the government, and issue a DOD "request" that Cisco do it, etc.. And at that point, you are competent enough to compete and win in the market place yourself, and competent people don't need to hack systems, because when they want them, they build them (or buy them). That's why you hardly ever see a competent act of terrorism: if they are competent enough, society will pay them more for playing the game than they could ever obtain by willfully refusing to play the game. > Or I come up with an RFC for DaveP, write a sample implementation, > demonstrate that it is more efficient than IP, go through the > standards process, get a few sites to run it, the router mfrs add code > for it, and now I can send packets without protocol values. Yes, you could do that. It would take at least as long to get deployed as IPv6. Which is why I suggested that you fix the FIN-WAIT-2 bug, while you were at it. 8-). > The point is, it's less of a law than gravity. Actually, it's not. Until there is an alternative to falling when you step off a cliff, falling is your only option: that's the law. After millions of years of evolution, and 12,000 years of civilization, and 6,000 years of recorded history, THEN you get the option of having a hang-glider hold you up. For DaveP, you will need to gain something that you cannot "brute force" this way: you will need willing cooperation. > Popping the stack here, email standards are difficult, but not > impossible, to change. Go to http://www.imc.org/ and start today, if you want us to have different laws. Just be sure those laws have the emergent properties you want the final system to exhibit, before you carve them in stone. 8-). > Gravity is extremely difficult to change, if not impossible. Email > standards are relatively new, Gravity is fairly everpresent in the > timeline of human endeavor. Some people would argue about gravity. It depends on what you mean when you say you are "defying" it, whther you meant the underlying fundamental principle, or whether you mean one of the emergent properties of the underlying fundamental principle. For example, it's possible to redesign the "ball in the falling elevator" experiment so that you *can* tell you are freefalling in a gravitational field, rather than being out in space. 8-). > Popping back to the original frame, it's in our power to change how > email works if we don't like how it works now. Complaining about how a > strange attractor of a willful disruptor (which actually comes into > being _because_ of how people are) is less useful than actually > figuring out how to make the email system better. Well, I was never aruing about that, per se, I was arguing about how to overlay a system in the context of an existing system, in order to damp the effects of the strange attractor. When we talk about SPAM countermeasures, like blacklisting, what we are talking about is active damping, in place of replacing the system with a system which has an emergent property of passive damping. You argue that active damping is evil; my response is "replace the system with one that contains passive damping as an emergent property, and we will stop active damping". It's incumbent upon you, as the person who wants to prevent the *natural and predictable* response to the strange attractor of tighter controls on identity and content, to provide an alternative in the form of systems engineering on the underlying transport system. Because as it stands, you aren't convincing anyone. > Finally, asserting that a particular technology _which we have control > over_ demands that people behave in a certain manner is naive and > borders on the dishonorable. We can change the technology to adapt > to how people behave and limit "damage". For any arbitrary value > of damage, this has a better chance of working than demanding > that people change their behavior. We have such a technology, already widely deployed in the world: we call it "government". If you don't like it, engineer a different system, and convince the world to deploy it, in place of the current system. But any feedback system, even a passive one "built into the rules" and operating "because that's the way things are" is going to "demand" people's behaviour conform to certain normative values within the tolerances of the system. Or the people exhibiting that behaviour *will be* damped. -- Terry To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-chat" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3E4A7B76.2D6856C6>