The Byzantine Trap

With all due respect to Steve DeAngelis, but his solution won't work. Nice attempt, but the philosophy is entirely wrong. As with many things of this level of complexity, if you get the philosophy wrong at the start, the solution is even worse than doing nothing.

Comments

[shloky]

Yeah, I thought the world was decentralizing.

Also, what happens when Oak Ridge gets hit?

[Michael Tanji]

After reading ES site literature and this article – and let me know if this is just me or not – but does it not all sound like very other mondo “software will save us all” project you’ve ever heard in your life? I’m not trying to be a d!ck or anything; either I just don’t get it this is just like every other grand idea that someone tried to translate into code (with predictable results). The ideas are all great and the sentiments are righteous, but as you suggest: what happens when something you didn’t code takes place? What helps you defeat garbage-in, garbage-out?
Closing the gap would be easy were it not for all the warlords and dismal environment and (necessarily) suspect people and all the other issues related to the planet’s garden spots. If you can’t polish the t#rd that is Iraq with a couple of Army divisions, the Marines and Air Force hanging around you expect to carry this off in places like Somalia? Who is going to back this? The body-shops like UN/IMF/WB that the software aims to replace? The private sector that is watching Bechtel pack its bags?

Is my career as a big thinker over before it started or am I just a jaded malcontent?

[scalefree]

This kind of reminds me of Bob Steele's plan to divert about half the military/intelligence budget to OSINT. As much as he's seen as the godfather of OSINT, there's just something about the phrase "open source" that he doesn't seem to get.

Centralized, all-inclusive systems are by their nature not resilient. Resilience comes from decentralization, redundancy, multiple competing strategies, multiple points of failure, etc. You can't elect yourself hub of it all, hubs have to prove their fitness over time.

Still, I bet they have lots of cool toys & smart guys making them do neat things. Lots of blinkies I bet.

[Critt Jarvis]

Scalefree opines: "Centralized, all-inclusive systems are by their nature not resilient. Resilience comes from decentralization, redundancy, multiple competing strategies, multiple points of failure, etc. You can't elect yourself hub of it all, hubs have to prove their fitness over time."

I agree, fitness has to be proven over time. As for centralization, I imagine IATGR as "a" hub, rather than "the" hub. If "a" hub, then a community of interest could be wherever two or more hubs are gathered.

Snip from Steve DeAngelis, blogging Self-organizing Rule Sets:

"When I talk about rule sets and standards as they relate to Development-in-a-Box, however, who makes those rule sets and how they are enforced is much less clear to people. This is especially the case since I do not envision a "governing body" that either oversees the process or enforces its implementation. Development-in-a-Box relies on the voluntary cooperation and collaboration of disparate groups; hence, the ambiguity about rules sets."

This [www.iatgr.org] could turn into a really interesting participatory experience, for those who choose to participate.

Of course, to start, it would be helpful if they had a blog, with comments enabled.

Hmmm... what else do they need that might make you feel invited to the party?

[deichmans]

I'd like to add to Critt's comment, which acknowledges a very valid kernel in John Robb's criticism (i.e., that we've all heard "software über alles" before and not seen it deliver). But the real question is not "Will this work right now with my fractured architecture?", but rather "Could this work?"

Robb makes the presumption that any philosophy of resilience is fatally flawed, but he does not articulate what that philosophy is.

This issue ultimately boils down to one environmental metric: is it "permissive" or "non-permissive"? We trust the traction control system and anti-lock braking systems on our cars to decide for us ("permissive" environments -- the rules are known, and the ice on the road is not actively trying to make us spin out). Can this simple kind of solution now be scaled to the more complex with a thinking/adapting/malicious adversary? ( q.v., the comments on Robb's continuation 'blog "The Byzantine Trap 2..." -- http://globalguerrillas.typepad.com/johnrobb/2006/11/the_byzantine_t_1.html -- which opines that any working complex system has been derived from a simple system that works.)

This is the crux of the issue, and one that I've grappled with for years as the "Chief Devil's Advocate" at U.S. Joint Forces Command. While I sincerely doubt we'll ever achieve true "predictiveness" on the battlefield, I do believe we can anticipate enemy actions and bolster our ability to absorb them. This is the true essence of "Resilience": The ability to continue providing goods and services, despite system perturbations.

vr/ shane

[John Robb]

Hmm. I contend that within exceedingly complex environments, the only true way to approach resilience is through decentralized processes. If you don't approach the problem from this perspective (a philosophy of system design), the complexity overwhelms you and you fall into a cycle of rapidly diminishing returns.

[zenpundit]

"I contend that within exceedingly complex environments, the only true way to approach resilience is through decentralized processes"

Often this ( decentralization for redundancy) will be the best strategy, no argument. But always ?

Consider that simplicity in system design and function might be another. Fewer variables to go wrong than with a widely distributed system. Technically it is not as resilient in the sense that there is no " back-up" but what about initial probability of failure ?

Secondly, John Shane, Critt - in terms of resilience in " noisy" environments, adaptiveness might be measured not simply by adding new rules to the rule-set in response to " noise" but in ruthlessly subtracting the merely less efficient rules on a systemic basis to make way for new, emerging, solutions from the bottom up.

What do you think ? You guys have the tech and engineering backgrounds -I'm just a poor historian ;o)

[John Robb]

Zen, the problem within the context that we are talking about (societal resilience), is that further specialization/centralization/sophistication/control isn't scalable nor does it provide positive ROIs. Of course, there is a solution: simplify the system's assumptions by building a decentralized platform. Once the framework/platform of a decentralized system is in place, then centralized systems can be rebuilt in ways that are both practical and beneficial. The political problem is that these decentralized frameworks don't have any political currency. They cut against historical sources of state power.

[Account Deleted]

John, why are you assuming that decenralization is analogous (or whatever is meant by "/") to specialization?

Also (and unrelatedly) could you give an example of a "decentralized platform"?

[scalefree]

I'll name two decentralized platforms for you: the blogosphere & bittorrent. Both harness the math of network effects to accomplish their goals of moving information in a (more or less) coherent fashion; implicitly in blogs & more explicitly in torrents. I'll even give you a third that just came to me, DNS.

As for resilience vs complexity, I'll point you to the work of Joseph Tainter. His book The Collapse of Complex Societies is a classic. You can find some good extracts online if you're too stingy to buy it, which I'd link to except Typepad ate my links. The upshot is that mathematically the energy requirements of a monolithic project on this scale doom it to failure.