Stephanie Forrest

http://www.cs.unm.edu/~forrest/

We recently discussed this popular article about her work ...

• http://www.networkworld. ... mune-system.html

In response Ian Brackenbury said . . . The article doesn't give away much re implementation and so I'm sceptical of the approach and, as you say, it's not the multicellular domain that Steve thinks is the pertinent analogy.

Take ?JavaScript for instance, it seems a tad self defeating if each machine interpreted ?JavaScript programs with unique semantics, yet ?JavaScript is an (increasingly) capable language and certainly rich enough that the effect of running a ?JavaScript program cannot be determined by inspection, only by running it. The same goes of course for Java and I don't see that altering the bytecode assignments would be very productive either.

So I assume this approach is about the running of connected systems where there is no authorised intentional downloading of code and therefore the issue is how at hardware level to make up for inadequately secure software operating environments. It might save rewriting the apps written in C/C++ for Windows. A slightly less constrained solution would perhaps allow some intentional code download but only using digital signing and sender authentication of all legitimate moving code - Stephanie's scheme would expect to catch any rogue injections of code.

It seems to me that the idea of sandboxing has been beneficial - if the entranceways through the sandbox walls are as carefully designed as Steve would have us design the interfaces into each "multicellular" computer then the effects of running code of arbitrary complexity inside a sandbox should have bounded harm, even though it isn't amenable by inspection to proving only intended "safe" consequences.

I wonder if the idea of sandboxing has analogues in the multicellular organism world. That might mean careful attention was paid not just to external interfaces outside the multicellular organism but between specialised components of the organism as well. (Steve might point to mitochondria perhaps) I still cleave hard to my view that it's not code per se that is the issue but that the language of the valid messages needs to be so simple and obvious that any conversation is immune from consequences unintended by the receiver.

By the way, the "simple and obvious" test of an interface language should be true for both Recipient (Steve's current focus) and the Sender. Is this possible I wonder - it's like a contract for service of a washing machine; it should be written in such a simple subset of English that both parties are sure there's no "small print" zinger in it. The service provider wouldn't want accidental commitment to lifetime servicing af all your washing machines for ever; you wouldn't want the contract to allow the servicer legal access to your house 24x7 - it works both ways, in my view. -- IanBrackenbury

Copy of a Massey University press release added by Ian Brackenbury (who distrusts the longevity of URLs to news on the WWW). Challenge to established view of eukaryote evolution: EukaryoteEvolution. If some of the persuasive power of Steve's argument was that evolution of single-celled to multiple-celled organisms gave a practical example of mechanisms that led to great reproductive and survival success then challenges to the current understanding of that evolution will be of interest.