% Thoughts: Docker, Sociology, Mathematical Models of Security, Word Choice, and Death % Michael Stone % July 16, 2014 1. Multi-container docker configuration is presently mired in complexity (examples: [confd](http://brianketelsen.com/2014/02/25/using-nginx-confd-and- docker-for-zero-downtime-web-updates/), [gantryd](https://github.com/DevTable/gantryd)) in part because docker makes it hard to deterministically assign addresses to containers. As a result, service- discovery must be done via dynamic control and agreement at "network run time"instead of at "network compile time". Potential solutions: a. [dnshash](http://dev.laptop.org/git/users/mstone/dnshash/plain/README) b. enable deterministic address assignment 2. Translation, "problematisation", "interessement", "enrolment", "mobilisation", "boundary objects", "data-information-knowledge-control", and friends: * ["translation"](https://bscw.uni-wuppertal.de/pub/nj_bscw.cgi/d8022008/Callon_SociologyTranslation.pdf) * ["boundary objects"](http://en.wikipedia.org/wiki/Boundary_object) * ["actor-network theory"](http://en.wikipedia.org/wiki/Actor%E2%80%93network_theory) * ["Reframing Organizations"](www.amazon.com/dp/0787987999) 3. Impossibility results, information flow control, non-interference, and inference control: * ["A Hundred Impossibility Proofs for Distributed Computing"](http://groups.csail.mit.edu/tds/papers/Lynch/podc89.pdf), Lynch, 1989 * ["Protecting Privacy using the Decentralized Label Model"](http://www.cs.cornell.edu/andru/papers/iflow-tosem.pdf), Myers & Liskov, 2000 * ["Security Policies and Security Models"](https://www.cs.purdue.edu/homes/ninghui/readings/AccessControl/goguen_meseguer_82.pdf), Goguen & Meseguer, 1982 * ["Unwinding and Inference Control"](http://www.computer.org/csdl/proceedings/sp/1984/0532/00/05320075.pdf), Goguen & Meseguer, 1984 4. Most of the world uses unfortunate definitions for words like "identity" and "authentication" but my current preferred definitions: * identity: a cluster of nyms * intuition: to me, an exemplary primitive ascription of identity is a claim like "one of the authors of message 23 is also a recipient of message 47". As a result, I view "identities" as being particularly interesting subgraphs -- e.g., those identified by interesting closure operators -- on the graph of "evidence linking nyms to messages". * authentication: a guarantee that, at the end of a protocol, [some principals agree on a mapping of symbols to values](http://web.cs.wpi.edu/~guttman/pubs/auth_tests_long.pdf) are way too surprising to reasonably be called "identity" or "authentication". (What to do!?) 5. Somehow, I've never found traditional explanations of death very satisfying. Thus: if "life" is usefully undertsood as a particularly interesting collection of emergent properties, then what would a [CAST analysis](http://psas.scripts.mit.edu/home/wp-content/uploads/2014/03/CAST_Accident_Analysis-1_PN.pdf) of some particular organism's death look like?