Monday, March 25, 2013
Sunday, January 06, 2013
- Convexity is easier to attain than knowledge (in the technical jargon, the "long-gamma" property)
- A "1/N" strategy is almost always best with convex strategies (the dispersion property)
- Serial optionality (the cliquet property)
- Nonnarrative Research (the optionality property)
- Theory is born from (convex) practice more often than the reverse (the nonteleological property)
- Premium for simplicity (the less-is-more property)
- Better cataloguing of negative results (the via negativa property)
In information security, this implies that the goal should be to ensure that catastrophic breaches are structurally impossible. Once you've assured this, you can drive the ongoing the ankle-biter attacks down to a dull roar level that can be tolerated indefinitely. Network engineers avoid catastrophic failure with techniques like carrier diversity. Unfortunately, they still mostly allow the Cisco vendor monopoly to continue.
System architects in other IT areas haven't learned this lesson, either. They still allow their entire enterprises to be dependent on "monocultures" of products from SAP and Peoplesoft and the like. Here the herd instinct predominates. "Nobody ever got fired for buying from IBM" -- if IBM, or ADP, or salesforce.com goes down, everyone else goes down too, and you can't be especially blamed. But if you had diversified, then your enterprise wouldn't have gone totally down, and you would be positioned to step in while your competitors were struggling or failing, and win big.
Wednesday, December 26, 2012
Sunday, October 21, 2012
- Inability to distinguish dominion from destruction. In the Biblical story of creation, when God cast Adam and Eve out of the Garden of Eden, he gave them authority over all the living things of the earth. Six thousand Biblical years ago, people didn't have the ability to do much more than protect their villages and livestock from large predators such as lions, wolves, and eagles with spears and arrows. Now we have barbed wire fences that span continents, massive farm tractors that can do the work of 500 horses, and heavy earth-moving equipment that we use to literally move mountains in order to obtain the coal within them. We can change entire ecosystems, and we do. Conservatives need to take their Lord's injunction far more seriously, and cast off childish attitudes that they are helpless against the might of natural forces. If I were God, I'd be asking "What have you done to my garden? You have killed thousands of kinds of animals that I took care of myself because Noah only had only one Ark, and you have turned vast regions into lifeless deserts, and you are planning to do even more. I have sent many prophets such as John Muir, but you have not listened. Woe be unto you!"
- Viewing all of nature as a store of resources to be exploited. Just like "the only good Indian is a dead Indian", the only good land is land that can be farmed or mined, preferably both. But mining takes preference, regardless of its destruction of agricultural capability. When I was young, my friends would go swimming in the nearby "stip pits" that had filled with water after they had been mined and abandoned without even restoring the topography back to its original gentle hills. Before the EPA and related legislation required mining companies to replace their tailings, you could drive on US Highway 40 for a hundred miles through Indiana and Ohio -- the best farmland in the world -- and see nothing but hundred-foot-high ridges of strip mining spoils, with the occasional giant excavator showing its masts above them. But before it was farmland, those Midwestern plains were tallgrass prairie harboring hundreds or thousands of different species of grasses and insects. Now that land is planted with genetically modified corn, soybeans, and wheat that is poisonous to insects, and cultivated with "no-till" methods that save fuel used for plowing by saturating the soil with herbicides, so that broadleaf weeds and prairie grasses cannot survive. The result is mile after mile of a single-species landscape that is held hostage to the patent-protected seed stocks of Monsanto and Pioneer Hi-Bred, and can be catastrophically wiped out by unplanned weather conditions or invasive, pesticide-resistant fungi or caterpillars. To the conservative, this is good, because it allows those companies to extract higher profit margins today by deferring the cost of damage to future generations.
- Nature is the ultimate outgroup. Conservatives are an exclusionist movement. They want everyone to think like them, and they spend a lot of time arguing about who is a true believer and who is, for example a "Republican in Name Only" and attempting to expel them from their group. One suspects that if the technique hadn't been invented by the Chinese Communists, that they would be using "self criticism meetings" in order to shape behavior. Religious groups with their affirmations of commitment serve a similar function in "separating the sheep from the goats", and driving all differences towards the core beliefs, regardless of merit. Nature, of course, was there first, and it cannot be controlled, directed or shaped. Whatever your religious or political doctrine, nature will not follow it. This must not be allowed. To the conservative ideologue, untamed nature cannot be permitted to have any legitimate status in the community.
Friday, September 07, 2012
- closed loop material
- open loop material
Saturday, July 07, 2012
I'm sure that in 1888, when the U.S. Census declared the American Frontier to be closed, and there was no more "unoccupied" land left to be taken by the white man, and the US was still in the chaos of Reconstruction from one of the worst civil wars in history, making what's going on in the Middle East now seem like child's play, people were reasonably asking whether the country could survive another 112 years like those that had occurred since 1776.
And the the millennium occurred and those 112 years had been survived with substantial success. The U.S. economy in 2012 with air conditioning, jet airliners, internet video, and electric automobiles, not to mention hedge funds and risk arbitrage, is very different from "civilization as we know it" in 1888.
I have no doubt that the US and its economy will be as different in the year 2248 as an economy of 50 states is different from that of 13 English colonies. There is no doubt that many politicians will continue to be venal, corrupt hacks, as they have been for the past 236 years, but they will probably still have been elected by a majority of voting citizens who will get what they asked for.
Saturday, June 30, 2012
I have a brief counter-commentary -- They're both wrong:
"Meet the new boss, same as the old boss." The author and his academic sources don't seem to notice the contradiction in what they're saying. That is, that democratic movements can't succeed unless they are undemocratically organized with a dictatorial head or junta to "knock heads together and get everybody to stick to a plan." Karl Marx believed that there would be a "dictatorship of the proletariat" which would fade away to produce true communism. The Soviet Union's dictatorship did indeed fade away, but it was followed by the pseudo-democratic autocracy of Vladimir Putin, not by communism. The Romans tried electing their "dictator" who would voluntarily step down after the wars were over, but that didn't last long, ending when Julius decided to call himself Caesar and become emperor rather than step down. It's not yet politically or academically respectable to say that all forms of government evolve to become dictatorships or monarchies, so we end up with incoherent articles like this one.
Face it, democracy is hard. It requires the people to elect representatives, not leaders. It requires the people to communicate thoughtfully with those representatives, and the representatives to reasonably and thoughtfully work with each other on common problems. When major political movements are based on the premise that negative campaign ads work better than constructive discussion, that cooperation is evil and that members of other political parties are traitors, democracy will continue to deteriorate.
Social media have the opportunity to bypass power-hungry leaders and allow the people to communicate directly with each other, making it possible for leaderless democratic movements to react and refocus more efficiently and rapidly than ever before, but their technical architecture with centralized software and servers makes them just as corruptible as the old fashioned political machines that used smoke-filled back rooms instead of giant server farms.
Saturday, April 21, 2012
The tried and true aphorism [is] that government isn't any good at picking winners. This isn't, by the way, a knock on government. No one is particularly good at picking winners. The problem for government is that while market-produced losers usually fail and go away, making room for winners, government-produced losers tend to stick around for a while, sucking resources away from potential winners. No one knows in advance whether something will work; government's failure is in its relative unwillingness to clear away the chaff.In economist-speak, the subsidies that Free Exchange describes are "Pigovian subsidies", the converse of the more well-known Pigovian taxes. The reason that this kind of tax works better than subsidies is that the lifecycle of an enterprise is asymmetrical -- the growth phase is much shorter than the decline phase, so the cumulative penalty of a tax during growth is less than the drain during decline, putting failing enterprises out of their misery earlier, while the cumulative effect of subsidies is reversed, having a smaller effect during growth while prolonging declines. If a government creates equal numbers of subsidies and taxes with equal rates, the total effect will be negative.
Friday, March 30, 2012
Computers and robots don't make mistakes, right? But there's a complementary saying: "To err is human, to really screw things up requires a computer." After all, the things were designed by error-prone humans.
Fast food restaurants have achieved much of their success by creating a product that can be considered "high quality" in that it is identical each and every time you go to one no matter where in the country, and almost in the world it is. Yet on my last visit to one of the top 3 franchises, they got 3 of 5 items in my order wrong, and while I was there they made errors in the orders for two more customers.
One of the reasons that Apple is such a powerhouse is that they have achieved a higher level of quality than their competitors can ever aspire to, even a decade or more after the infamous "blue screen of death" was common.
But a monolithic ecosystem of total control is not the only path to quality. Most of the web runs on the Linux OS and Apache web server, which are both completely cooperative, transparent, loosely coordinated enterprises, and achieves higher quality than its closed, commercial competitors.
Engineers have another slogan: "we can build it for you fast, cheap, or good. Choose any two." You could survey people asking whether they've become used to fast and cheap, and "good" has become impossible.
Luxury is a surface characteristic any more. The luxury smartphone doesn't have any better software or give you any better sound than the iPhone that millions of people carry, though it may come in a gold-plated case. Even Bill Gates runs Microsoft Windows on his PC.
My taxes were almost lost when my tax peparer's PC crashed at the end of tediously entering all the data. He wasn't sure that it had made a backup for him; he hadn't bothered to check that any kinds of backups were ever made.
My Toyota Prius has the same navigation software as a Lexus, although the Lexus has a somewhat bigger screen. Both systems have the same bugs and usability problems.
Toyota has a well-deserved reputation for quality, but they can't deliver an operator's manual that correctly describes how the hand controls relate to the headlight settings. As I was sitting with the "finance manager" at the dealer completing my purchase, his computer crashed and was unable to print some of the government forms - we had to move to another office to finish all the paperwork.
Now, what might happen to General Motors dealers' ability close a sale with correct pricing and product option information when their new Chief Information Officer is known for decimating the Information Technology Department at his previous job, firing all the high-salary, experienced veterans and replacing them with low-wage workers offshore? How can quality be maintained in a regime whose goal is rapid delivery at unprecedentedly low cost?
Quality issues have been an issue with systems since before Capt. Grace Murray Hopper found the first insect in the backplane of a mainframe. In my job I'm dependent on people doing high quality system design and operating those systems reliably while making changes to them, effectively rebuilding the metaphorical airplane while in flight.
Today it seemed that all my problems were quality-related. To top it off, when I went to my e-book reader's online store to buy "The Checklist Manifesto" for some weekend reading, the shopping cart function wouldn't work. Argh.
Saturday, February 25, 2012
This same technique could be used by owners of patented computer algorithms -- let your algorithm escape from your licensing control and become incorporated into a computer virus or worm, then demand royalties from the owners of the infected computers or sue them for infringement.
I have no idea whether this generalization of Monsanto's trolling method is patentable as a business process. Nor have I read the filings in the farmers' lawsuit, so I don't know whether they contain any hints that any participants in the case are aware of the generalizability of the method. But if they aren't, here's a statement that I believe this is an obvious generalization, and its obviousness should be grounds for invalidation of any attempt to patent the trolling method.
Wednesday, February 22, 2012
It's easy for a biologist to imagine that social and economic progress follows a logistic curve that starts out exponential but flattens off as resource limits are approached. It's also the case that after the easy problems are solved, the remaining ones become exponentially more difficult, producing the same slowdowns, though without the hard upper bound. The social difficulty is that, as Thiel observes, our political systems are built on the promise of never-ending growth. You can't get re-elected by promising that there are not going to be any more free chickens for every pot.
More compute power is an essential prerequisite for getting out of this trap. With enough bandwidth and large enough displays and enough compute power to drive them, "as good as being there" can become a reality and the limits to material resources and the costs of transporting those resources can become effectively non-issues. However, better, more reliable, more trustworthy software is also a prerequisite. The complexity mountain is a problem for software, too.
Wednesday, October 12, 2011
Sunday, August 21, 2011
- Something must be done
- This is something
- Therefore, it must be done
Friday, July 08, 2011
Wednesday, June 29, 2011
Monday, May 30, 2011
There's the paranoid approach: "however much security you have, it's not enough" This is encouraged by vendors of security products and services, who want you to buy, buy, buy, and don't care if you're spending your money effectively. It's functionally equivalent to the "priceless assets" approach: "if your assets are infinitely valuable, anything less than an infinite amount of spending on security is inadequate." This approach is deeply baked into the security industry due to its origins in military security, where the asset value is the entire country.
Then there's the loss-management approach. This is based on the notion that losses can be predicted, and controls can be assessed for their effectiveness in mitigating those losses. This is the only approach that that provides a principled basis for a budget less than "all the money you have". But how do you manage effectiveness in a principled way, when vendors are motivated to tell you "trust me, it really works great!" and hide any weakenesses that their product or service may have until it's too late for you. Third party certifications such as Common Criteria protection profiles ensure a baseline of effectiveness, but the CC certification hierarchy doesn't distinguish levels of effectiveness - the distinguishes trustworthiness of achievement of the baseline. A product certified at EAL4 may may be no more effective than one certified at EAL2.
Assessment of effectriveness is problematic prospectively, but it can be assessed retrospectively: simply add up the losses actually experienced with a given configuration of controls. That is, if you are unable to develop a credible estimate of annual loss expectancy, use historical data for measured annual losses. That is, ALE = MAL.
Now apply the principle of not spending more than the value of the asset to your annual budget. You have observed MAL, so you can say the annual security expenses shouldn't exceed that value. SE =< MAL.
In an environment where threats & assets cannot be effectively and reliably estimated, security expenses will approach an equilibrium with security losses. This is not good news for participants with assets that are protected by the laws of macroeconomics, such as consumers in a free-market economy whose personally identifiable information is somewhere out there in the cloud.
Saturday, May 28, 2011
Wednesday, May 18, 2011
By Richard Hofstadter
Harper’s Magazine, November 1964, pp. 77-86.
Hofstadter was a famous professor of political science at the school where I was an undergraduate, though I never took any of his courses. This article is one of the reasons for his fame. The paranoid style is evident to any careful observer of politics, but this puts it in a broader context. No you're not imagining it, they really are crazy, and they've been that way for a long time.;
Friday, April 22, 2011
Sunday, March 27, 2011
Not to mention his fully fabbed Rule 110 cellular automaton, with a few parts missing, oops. Rule 110 CA's are also universal, with a nice scandal to go along with their discovery.
Reminds me of one of those naive geniuses that pop up regularly in the SF literature.
Monday, February 21, 2011
- sustainable in the sense of lasting at least as long into the future as civilization has extended into the past, some 3000-5000 years.
- stopping the decline in biological species diversity. This can occur either by forestalling the extinction of existing species, or by increasing the rate of appearance of new ones. Currently we're out of equilibrium by at least 10,000 to one.
- Stopping the increase of carbon dioxide in the atmosphere, and secondarily other pollutants itself. The days of "the solution to pollution is dilution" are long gone. Mark Z. Jacobson's GATOR model is an example of the state of the art in this area.
- Transformation of the global energy economy to sustainable sources. Jacobson and Mark Delucci have concluded that it is technically feasible to transform global energy sources to wind, water and solar within 20-40 years. They are of course wildly optimistic since neither the political will nor the economic resources are available.
- Thus integrating economic models into this transformation is necessary. Things are unlikely to change in ways that are unprofitable; causing economies to collapse by raising taxes to unsustainable levels in order to fund energy projects doesn't do anyone any good. Cyclic booms and collapses don't count as "sustainable" even if their long-term average is nonzero.
- We suspect that it will somehow be necessary to decouple the material economy from the nonmaterial value chain. Many material resources are bounded, but billionaires cannot personally consume all their wealth; it's just places they don't have time to go to and money they don't have time to spend. We would like to know whether a level of health and comfort equivalent to a first-world country in the year 2000 can be achievable for most everyone in the world using market economies.
Monday, December 20, 2010
Fiddledeedee 555-5555-55555-5552. That's fifteen 5's and a two.
I think it must be related to an observation that many weather prognosticators and climate change skeptics' declarations cannot be distinguished from the results of numerological computations on facts about clowns.
Masters writings of course are the very antithesis of this approach -- they're as scientific, coherent and data-based as it is possible to get.
Friday, October 15, 2010
The New York Times doesn't. An article by David Segal concludes that economics isn't really trying to be successful -- people are just too complicated. He quotes Duke University professor and specialist in behavioral economics Dan Ariely, who says "...the economy is a hugely complex problem. So we either simplify the problem and offer a solution, or embrace the complexity and do nothing.” Or as I say sarcastically, "if at first you don't succeed, give up."
Sunday, October 10, 2010
Saturday, October 02, 2010
Tuesday, September 28, 2010
Monday, September 06, 2010
Saturday, September 04, 2010
- it never happened
- even if it happened in the past, it's not happening now
- even if it's happening now, it's not due to anything we did
- even if it is due to something we did there's nothing that can be done about it
- even if there's something that can be done about it, it shouldn't be done for other reasons
- even if it is due to something we did, it wasn't with malicious intentions and we shouldn't be held responsible
- even if something should be done, we shouldn't have to pay, somebody else should
- even if we ought to pay for the fix, paying will consume all of our profits and we'll go bankrupt and then somebody else will have to pay anyway
- even if we won't go bankrupt, our profits will be reduced, and this is bad for the country if not for the world
- solving the problem is revenue-neutral, we could get a lot of good press and "brand reputation" if we fixed it
- hey, we could increase our profits if we really fixed this problem
Saturday, July 24, 2010
When the EPA first made its finding that the climate impact of anthropogenic CO2 and 5 other greenhouse gases endangers the health of U.S. citizens (the "CAA endangerment finding"), the Obama administration made it clear that if Congress didn't produce legislation, the EPA would act unilaterally. No deficit-reducing taxes, no free-market cap and trade framework, simply a flat limit on emissions, just like benzene, ozone and other pollutants. "Please don't throw me into the briar patch!"
Friday, December 11, 2009
How secure is ODBC?
Any ODBC sniffer will be able to trace everything from an ODBC perspective. This includes data, usernames, passwords etc. However, if you are using an ODBC driver that provides encryption, you can increase your level of security.
Since any front-end tool can effectively connect to and modify your databases, you need to enforce security at the server level.
On the other hand, if you use TCP/IP, ODBC security should be the least of your concerns!
It should be massively embarrassing to every security professional that the basic rule of never transmitting or storing passwords in clear text still doesn't have a standard, default implementation even now, many years after the first ODBC specification was published in 1992. The fact that ODBC is really an API, and not a network protocol, and that it was created for a non-networked environment where communication between the client process and the DBMS would take occur in the within-system interprocess communication framework, via OS traps using shared memory or intrasystem messages where security can be rigorously enforced, instead of the modern environment where database client and server processes run on different computers with an open, possibly hostile network in between, is not really an excuse. Vendors have had seventeen years to work this out!
Some ODBC libraries do support SSL session encryption, and if you encrypt everything, then passwords get encrypted too. But passwords should be encrypted always and everywhere. If every OS was able to figure out that this is required decades ago, DBMS products should be able to figure it out, too.
Monday, December 07, 2009
While ownership of the name is good for that company, it's bad for the industry, which has to use some other, less felicitous term, such as HCNG, which is used by NREL. DOE’s Advanced Vehicle Testing Activity (AVTA) spells it H/CNG, and has vehicles using 15%, 30%, 50% and 100% hydrogen.
Sunday, November 29, 2009
A few weeks ago, Science magazine, one of the most prestigious general-readership journals (if you can call a polymath scientist a "general reader") published a short article in its perspective section by two of the most eminent computer engineering researchers in the US, William Wulf and Anita Jones, about computer security, titled "Reflections on Cybersecurity".Their summary is almost accurate "Cyberspace is less secure than it was 40 years ago. That is not to say that no progress has been made—cryptography is much better, for example. But more vital information is accessible on networked computers, and the consequences of intrusion can therefore be much higher. A fresh approach is needed if the situation is to improve materially." And their discussion, behind a membership barrier or a typically outrageous $15.00/day per article pay-per-view fee, is generally correct. They list a number of ways that security goes wrong even with the best designs and the best methods.
Their error is in their conclusion, that public key cryptography is the miracle cure: "we conjecture that by providing just a way of accessing the public key of an object, one could build an arbitrary end-to-end security policy." Yes you can probably build an arbitrary end-to-end security policy, but in my experience with public key infrastructures, it will be intractably complex, in the technical sense of being NP-hard to administer in all but trivial usage structures. This is the same kind of error that occurs in real life with role-based access control schemes: for naturally occurring organizations rather than artificial examples, you quickly end up with more roles than people, and the system, though elegant, costs more to operate and administer than the messy environment that you started with.
Any system with crystalline simplicity such as the one that Wulf and Jones are looking for will have the brittleness of crystals, too. Strike it at just the right angle and it will fail disastrously. They have failed to recognize the key design decision by Tim Berners-Lee that made the World Wide Web scale so remarkably. Unlike nearly all previous hypertext systems, the WWW does not automatically create backlinks with every forward link, and it doesn't automatically update links with their targets change or go away completely. The Web expects errors and deals with them routinely. Even the very advanced semantic web, which is otherwise little more than a type system for XML objects, expects to see uncomputable type specifications and deals with them routinely.
If academic researchers want to make significant advances in security, they need to come to grips with the notion of "robustness" and not confuse it with "simplicity", which although it is very similar in that simple systems are often easy to make robust, they're not the same. Two of the most robust systems we understand, the immune system and the behavioral programming of the nervous system, are also among the most complex systems known.
Friday, October 23, 2009
No principled economist should be for nuclear energy, because its costs are dominated by serious aspects with extremely long tailed statistical distributions. Unlike chemicals such as PCBs where the cost of projects such as the cleanup of sediments in the Hudson River is merely unimaginably huge, there has never been a cleanup of a nuclear site so successful that it’s now suitable for residential use.
Other chemical disasters also have infinite costs — consider the permanent loss of the entire town of Times Beach, Missouri due to dioxin contamination. It’s also true that the costs associated with coal tailings and other mining wastes have equally long tails. Picher, Oklahoma is being abandoned due to mountains of toxic tin mine tailings that cannot be cleaned up at a cost less than the total value of the town.
We cannot base a permanent energy economy on extraction-based activities that cause progressive, permanent damage to the environment — sooner or later we’ll end up with all of the environment contaminated, and we’ll have no good places left for ourselves. If you like nuclear energy, we already have a wonderful source of fusion energy that produces far more power than we’ve been able to capture so far, and it keeps its waste to itself, at a safe distance of 93 million miles. Photovoltaic, solar thermal, wind, hydro, and wave energy produce no toxic waste needing cleanup after the plants have completed their lifespans. Not to mention photolysis of water to produce hydrogen, which has a nice promise to make a chemical fuel in home power plants for people who have an emotional need for a viciously roaring internal combustion engine in their car rather than a meekly quiet electric motor. But solar hydrogen technology is much less farther along than the other renewable ones.
Natural gas is a useful low-carbon fuel, but it can only be a transitional stage to a fully sustainable energy economy.
Tuesday, September 15, 2009
The U.S. Energy Secretary, Paul Chu, has put the government on a path to a renewable, carbon-free energy ecosystem that is based on electricity and battery storage for stationary and short-distance transportation, and biofuels for long-distance transportation. This is a perfectly valid path but it's not the only one.
H.R.1622 was passed unanimously by the House and referred to the Senate Energy Committee on July 21. This bill directs the energy secretary to implement a 5-year program to enhance the capabilities of Natural Gas Vehicles (NGV) in 12 areas, including fuel storage, fueling stations and NGV-electric hybrids. These capabilities are a necessary next step, but they don't provide big picture that gets us to where we need to go. Here's a sketch of a path that does. There's a lot more to this picture than there is space for here. The National Renewable Energy Laboratory has done a lot of heavy lifting in this area.
- Expand interstate infrastructure for Compressed Natural Gas transportation, driven by demand from long-haul truck lines and by supply pressure from natural gas producing companies
- Develop capability of CNG motors, based on demand from trucking companies
- Provide CNG motors in autos, based on fuel-management technology developed for trucks. Just like diesel fuel, CNG cars can drive up to the truck pumps at the fuel station. Home fuel stations become viable for those homes that have gas heat.
- Deploy hydrogen-enhanced "Hythane" fuel. Hydrogen can be obtained by steam reformation of methane with carbon capture, or by direct production of hydrogen from water
- Develop "Hy-flex" engines that can run on any blend of hydrogen and methane from 100% CNG to 100% hydrogen. At this point pure hydrogen fuel stations become a technically viable proposition.
- Prohibit pure CNG
- Progressively reduce the allowed proportion of methane in Hythane fuel.
- Allowed proportion reaches 0%, prohibiting methane in compressed-gas fuel. Done!
Saturday, August 08, 2009
Interesting on its own, the argument also applies to IT risk management. CIOs like to simplify their systems, for many good reasons, including security reasons. The farther the system gets from being analyzable by the security staff, the more likely it is that it will contain a critical vulnerability that isn't being adequately addressed.
But they need to be sure that they don't simplify too much. We all know the maxim about not "putting all your eggs in one basket." CIO's like to say "we're an XX shop" where XX is IBM or Windows or SAP, but whenever they do this they're admitting that they're not only at the mercy of that vendor but also at the mercy of any cybercriminal who holds an undisclosed zero-day exploit. If an application or infrastructure component is so essential to the business that if it went down the business would also have to shut down, then that application or component probably needs to be partitioned, modularized, and diversified so that any single failure is not catastrophic.
Sunday, July 26, 2009
It's shocking, although I have to say not really surprising, that so much of the discussion in the debate about restructuring our healthcare system is about how to maintain the profits of the insurance companies at the expense of the health of U.S. citizens.
The other perversity of the current system is the fee-for-service model, that pays more for delivery of more procedures, regardless of whether they actually do any good for the patient.
The original vision for Health Maintenance Organizations was that they could reduce costs by keeping their subscribers healthy. Healthy people don't need treatment as often, so by providing programs that keep subscribers from getting sick, HMO's could reduce the amount of money they would spend on treatments. But they discovered that prevention programs have overhead -- they actually had to engage with their subscribers regularly, and convincing subscribers to stop doing unhealthy things and start doing healthy things was complicated and took work. It was much easier to simply deny care when they got ill, or better yet exclude people who were likely to get sick in the first place. If your HMO only accepts healthy subscribers, payments for treatments are low and their subscriber fees are mostly profit. So HMO's became care-denial organizations. This acted to counterbalance the motivations for unnecessary treatments, but it didn't do anything to keep patients healthy.
In order for the United States to have a healthcare system that promotes the health of citizens instead of working against them, we have to identify those portions of the system that are incentivized to work against the interests of the end-users and either reverse those incentives or eliminate those portions entirely. I don't know of a structure that does that other than a government-administered single-payer system. Yes, government is inefficient, but it could hardly be more inefficient than the current system that is full of middlemen and where every insurance company has its own unique set of forms for doctors to struggle with when they should be focusing on their patients, and the "statement of benefits" from the insurance company has 3 different prices for every line item.
Tuesday, June 02, 2009
Tuesday, May 19, 2009
Other risk or threat taxonomies can be found in:
- U.S. NIST SP 800-30 "Risk Management Guide for Information Technology Systems"
- SANS has a "What works" poster series that was organized by threat a few years ago. Unfortunately that perspective is gone from the latest version.
- recreational vandals: generally low skill level, looking for notoriety, will attack public-facing services but not destroy the ability of the site to show his "greatness".
- cyber-theives: looking for money, will extract credit card and other information that can be sold or used directly to generate cash. Will generally not do permanent damange so that he can come back later for more.
- cyber-extortionists: looking for weaknesses that can stop the service from fulfilling its function, so that he can threaten to do it again if the victim doesn't pay up
- cyber-spies: looking to get in and out without discovery. May be supported by governments and may have high skill levels and large amounts of resources available for cracking passwords and encryption keys
- cyberwarriors: attempting to do as much damage as possible. Will attack infrastructure for business continuity and data integrity.
Saturday, April 18, 2009
- What is fragile should break early while it is still small. Computerized systems always break; they need to be built so that any component, including the hardware and the OS, will not cause the system to fail if every instance of that component fails.
- No socialisation of losses and privatisation of gains. We haven't had a case where a computer systems needed a government bailout. Let's hope we never do.
- People who were driving a school bus blindfolded (and crashed it) should never be given a new bus. PCI and HIPAA penalties for data breaches need to be much more severe than the slaps on the wrists that are given these days.
- Do not let someone making an “incentive” bonus manage a nuclear plant – or your financial risks. CISO's should never report to the CIO. CIO's are paid to reduce IT costs; if they can do so by ignoring risks, they will.
- Counter-balance complexity with simplicity. Information systems are the most complex systems in any enterprise. Every time some local solution is added in because it's too hard to make a global change, risk increases.
- Do not give children sticks of dynamite, even if they come with a warning. While IT users may be system admins of their PCs at home, they should not be given that privilege over the systems they use at work.
- Only Ponzi schemes should depend on confidence. Governments should never need to “restore confidence”. If your IT systems are so complex that their risk can't be analyzed by different members of your security staff and yield the same risk results, you can't manage their risks consistently.
- Do not give an addict more drugs if he has withdrawal pains. Buying more security products does not often produce greater security.
- Citizens should not depend on financial assets or fallible “expert” advice for their retirement. If a security "consultant" uses some proprietary method that he can't teach to a company's security staff, he's likely to be making it up as he goes along.
- Make an omelette with the broken eggs. Don't remediate security weaknesses by patching on more controls, redesign the systems so that they are naturally secure.
Saturday, March 28, 2009
Evolution by natural selection is a natural phenomenon with the same status as heat flow, which has its own Laws of Thermodynamics. It's a statistical statement about aggregate properties of groups of individuals, which in the case of thermodynamics are atoms, and in the case of evolution are biological organisms. Once it's understood clearly and carefully, what was originally an empirical generalization turns out to be a mathematical truth as incontrovertible as the fact that 2+3 is greater than either 2 or 3.
- The first law of evolution: when entities that have inheritable traits exist in an environment where some of those traits make them less effective at reproducing themselves than entities with other traits, then evolution occurs.
- The second law of evolution: when one subgroup, or population, of evolving entities becomes separated from another subgroup of what was that same population, they will evolve into varieties that cannot interbreed with each other, creating different species.
- The third law of evolution: the complexity of organisms in an ecosystem tends to increase, because complexity has a lower bound, while it has no upper bound. That is, "there's always room at the top" of the complexity spectrum.
The laws of evolution apply to any entity that follows the first law, whether they are biological organisms, cultural memes, or data structures in an evolutionary algorithm in a computer.
Charles Darwin's great achievement was the discovery of the principles of evolution by examination of the fossil record and other sources. To the politically-minded, Darwinism is the recognition that the fossil record shows how evolution occurred in biological organisms.
In the 150 years since the publication of The Origin of Species, the theory of evolution has itself evolved, into a "modern synthesis" that is 60 years old now, incorporating molecular biology and population genetics. This theory (Huxleyism if you have to ideologize) recognizes that errors in DNA replication and recombination will lead to evolution, regardless of any evidence in the fossil record.
The latest features of evolutionary theory, still in progress under the banner of a wierd name, evo-devo, are the incorporation of developmental lifecycles into the organization of the traits that natural selection acts upon.
Update: I probably ought to mention, since I cite the laws of thermodynamics, that life occurs in an open system, on the slopes of entropy gradients, not in the closed system that the second law of thermodynamics applies to.
Also, Christopher Hitchens has a commentary about the Texas Board of Education decision. Although always entertaining, Hitchens doesn't actually add much light to the debate.
Sunday, March 15, 2009
You can look at economic systems in the same way. There are a few industries that constitute primary production, and the rest of us in other industries are all standing on their shoulders. Only primary industries produce real value that exists independently of some subjective judgement. If you are trying to design a monetary system that is immune to speculative fluctuations, you want the currency to be tied to a value that is objectively constant as possible.
The primary economic sectors are
- information technology
Many people would want to include real estate in that list, since other than waterfront landfills, they ain't makin' any more of it, as Will Rogers said. But we have seen a real estate bubble collapse recently, and there's no reason that it can't happen again. Real estate has a stable value only if you use it yourself. Real estate that is not owner-occupied free and clear is speculative by somebody.
Nevertheless, parts of the energy sector are set up for an investment bubble, as Eric Janszen wrote in an article for Harper's titled The next bubble: Priming the markets for tomorrows big crash. Once we're out of that bubble, we should have a sustainable economic foundation. There will be other bubbles, but they will be riding on that foundation, and investors who confine their focus to foundational sectors should be able to ride them out without too much damage.
We may discuss the sustainability of each of the primary sectors in future posts.
Saturday, March 14, 2009
Saturday, February 28, 2009
In his book "Consciousness Explained", philosopher Daniel Dennett (do we have to wait until he's dead to call him a "great philosopher"?) argues that the notion of "one mind" is an illusion, and that even as individual persons we have many minds. I think this goes a bit far, but it underscores that there's no neurologically justified way to look inside the box of our heads and tell how many minds are really there, as long as the mind that does the talking mistakenly says "I'm the only one". Nobody has a clue as to how to look at a neural circuit diagram and tell whether it can or does sustain consciousness.
Well, okay, here's what may be a clue: a unified consciousness is possible only in a system where the interconnect bandwidth between the processing elements exceeds the sensory bandwith between the respective processing elements and the external world.
Short of magic, ESP and telepathy, the one-mind options are physically impossible.
Even in science fiction, where one of the most creatively designed alien characters ever is a character in Vernor Vinge's A Fire Upon the Deep -- a group mind whose component elements look like puppies, but create a unified consciousness joined by ultrasonic sensors on their backs. Even with holographic multiplexing akin to the principles of phased-array radars, which doesn't seem to be what Vinge was thinking of, the bandwidth obtainable from atmospherically-transmitted ultrasound doesn't compare to the bandwidth obtainable in even a small fiber tract of neuronal axons and dendrites. Consider a simpler argument for bandwidth limitation: birdsong has much of its information in ultrasonic frequencies -- why are birds not telepathic with other birds of the same species?
Friday, February 13, 2009
This effect is well-known to users of eBay, Amazon, and other online marketplaces that incorporate reputation systems.
Sudipta Basua, John Dickhautb, Gary Hechtc, Kristy Towryc and Gregory Waymirec (2009) Recordkeeping alters economic history by promoting reciprocity. PNAS January 27, 2009 vol. 106 no. 4 1009-1014. Open Access article at http://www.pnas.org/content/106/4/1009.full
Saturday, January 31, 2009
.. software testing alone has limited effectiveness -- the average defect detection rate is only 25 percent for unit testing, 35 percent for function testing, and 45 percent for integration testing. In contrast, the average effectiveness of design and code inspections are 55 and 60 percent. Case studies of review results have been impressive:
- In a software-maintenance organization, 55 percent of one-line maintenance changes were in error before code reviews were introduced. After reviews were introduced, only 2 percent of the changes were in error. When all changes were considered, 95 percent were correct the first time after reviews were introduced. Before reviews were introduced, under 20 percent were correct the first time.
- In a group of 11 programs developed by the same group of people, the first 5 were developed without reviews. The remaining 6 were developed with reviews. After all the programs were released to production, the first 5 had an average of 4.5 errors per 100 lines of code. The 6 that had been inspected had an average of only 0.82 errors per 100. Reviews cut the errors by over 80 percent.
- The Aetna Insurance Company found 82 percent of the errors in a program by using inspections and was able to decrease its development resources by 20 percent.
- IBM's 500,000 line Orbit project used 11 levels of inspections. It was delivered early and had only about 1 percent of the errors that would normally be expected.
- A study of an organization at AT&T with more than 200 people reported a 14 percent increase in productivity and a 90 percent decrease in defects after the organization introduced reviews.
- Jet Propulsion Laboratories estimates that it saves about $25,000 per inspection by finding and fixing defects at an early stage.
Saturday, January 24, 2009
This action is inevitably controversial and led to the story being one of the "most commented" on the Chronicle's website. While they present the spectrum of good, bad, and ugly views that one would expect in the newspaper of a large southern city, one of the commenters made a very good summary of the scientific method. [I've added some formatting to improve the clarity.]
Scientists often get so caught up in the marvelous organization of nature that they forget the basics. Science is simple:Hey, that's five steps plus "repeat", one fewer than six days of creation and rest on the seventh. Which method is simpler?
Anything else isn't science, it's fiction or religion and should be taught in those classes, not science classes.
- look at some facts
- create a model of the facts
- use the model to make a prediction
- test the prediction
- revise the model to account for the differences between the model and the facts
Evolution is the best model we have for the organization and development of biological organisms. All the others are losers in the scientific contest. I have no problems with teaching the weaknesses in evolution as long as the weaknesses of all the competitors are taught, too.
Consider "creation science" or "intelligent design". All their so-called "facts" have been shown to be mistakes or frauds. Their advocates' theories are full of logical errors, they don't make predictions that aren't obviously wrong, and they don't ever go out into the field and test anything against what their God is actually doing right now. It ain't science, it's propaganda.
Wednesday, January 21, 2009
Arthur at the Emergent Chaos blog attempts to continue the debate on the viability of "risk management" as an approach to computer system security, based on the comment from Financial Cryptography that "risk management is a dead duck". [FC's home link is to itself in in SSL, but uses a self-signed certificate that neither Firefox nor IE7 want to trust. And with good reason. If a so-called security website can't get its own security setup correctly, can you take its content seriously?] In any case, the basic point of risk management as an illusion that practitioners should not pretend to be following seriously is a good one. Here are some of the reasons why risk management is more like the city floating in the sky in the distance across the blazing desert.
Sunday, January 18, 2009
This is really an intrinsic property of concentrations of humans, and the correct answer is not simply to design less stressful city plans with good transportation infrastructures so that you have fewer traffic jams, and mixed-use complexes so that you don't have to carry your groceries on the subway. These are stopgaps that make a bad situation less bad. The right answer is to get out of town, i.e. suburbia. People know this, even if architects don't.
Sunday, January 11, 2009
Sunday, January 04, 2009
Risk Mismangement by Joe Nocera is about the Value at Risk method for assigning a single number to a large collection of financial situations, at http://www.nytimes.com/2009/01/04/magazine/04risk-t.html?em
The End of the Financial World as We Know It is actually a two parter by Michael Lewis and David Einhorn, first at http://www.nytimes.com/2009/01/04/opinion/04lewiseinhorn.html?em and then, titled How to Repair a Broken Financial World, at http://www.nytimes.com/2009/01/04/opinion/04lewiseinhornb.html?em
I especially like the suggestion "Another good solution to the too-big-to-fail problem is to break up any institution that becomes too big to fail." When an enterprise is subject to catastrophic risks that put its existence in jeopardy, the only real remedy is to restructure the enterprise so that those risks are not catastrophic. That goes for the enterprises that are national and international financial systems, too.
Sunday, December 28, 2008
This is really the most important question in security: how much security is enough? If we're ever to get beyond the self-serving answer "you can never have too much security", this has to change. There has to be a framework able to distinguish "too much" from "not enough".
Sunday, December 07, 2008
There are other environmental consequences in store if these small power sources become pervasive. Before Edison and Tesla invented electricity there were thousands of little dams all over New England producing power to run the mills -- their traces can still be seen today. The environmental effects were serious, not some tiny meaningless snail. Among other effects, they destroyed the Atlantic salmon industry, put fishermen out of business and eliminated a source of healthy food.
But there are many rivers that already have many small dams to provide flat water for barge transportation. The McClellan-Kerr Arkansas River navigation system is one example, extending for 445 miles from Oklahoma to the Mississippi River, with 18 locks and dams. Power turbines could easily be added onto them without further damage to the environment.
Monday, October 13, 2008
We eagerly romp with [Taleb] through the follies of confirmation bias (our tendency to reaffirm our beliefs rather than contradict them), narrative fallacy (our weakness for compelling stories), silent evidence (our failure to account for what we don't see), ludic fallacy (our willingness to oversimplify and take games or models too seriously), and epistemic arrogance (our habit of overestimating our knowledge and underestimating our ignorance).