Turing Compliant?

[Alex]

No, it doesn't add anything new.  It might someday permit massive
parallelism, but the computations you could perform could be done on a
traditional computer in a formally equivalent way.  For instance,
roughly speaking, the idea for solving the Hamiltonian graph problem
this way involves encoding every possible path through the graph on a
separate strand of DNA, throwing them all in a beaker, and then doing
chemistry on the mixture that weeds out all but the strands
corresponding to a solution.  This is formally equivalent to just
iterating over the possible paths, testing each to see if it's
Hamiltonian.

Noone has got it to work on a really computationally hard problem yet,
AFAIK.  The Hamiltonian graph that was "solved" with DNA computing had
seven vertices, I think.  At any rate, you could pick out a Hamiltonian
path by eye in less than a quarter hour, for sure.  It took a week of
lab work to do it with DNA computing.

One problem with DNA computing is that as your solution space increases,
the volume of the liquid you store the DNA strands in has to increase,
too.  If the strands are too concentrated, you run the risk of importune
reactions between them.  At some point, it will also get difficult to be
sure that you've manufactured every candidate solution -- for the
Hamiltonian path solution, they made strands of DNA that represented the
graph's vertices, and arranged them in such a way that two strands could
anneal if and only if the corresponding vertices had an edge between
them.  Then they threw those in the beaker and waited long enough to let
all possible paths form.  As the size of the graph grows, "long enough"
can get very, very long.

Alex.

-- 
If cars were like computers, they would go 300 m.p.h. and get a hundred
miles to the gallon and cost $50. Except that twice a month someone a
thousand miles away would be able to blow up the car, killing everyone
nearby.