Re: choosing powerful primitives for a simplified computing system

--===============0442470378==
Content-Type: multipart/alternative; boundary=bcaec54fbaf66b75e904c52e865d

--bcaec54fbaf66b75e904c52e865d
Content-Type: text/plain; charset=ISO-8859-1

You give compression as an example of a powerful computation primitive.

Are you interested in the similarity between Matt Mahoney's ZPAQ virtual
machine
http://mattmahoney.net/dc/zpaq.html
which implements predictive compression by mixing a swarm of predictors,
each expert at a different domain.
And Miller and Drexler's idea of "Agoric Computing"
http://e-drexler.com/d/09/00/AgoricsPapers/agoricpapers.html
where they propose routinely combining untrusted agents into reliable
services.

I think your focus has been on eliminating the software complexity caused
by history.
But there's another slightly different focus - reducing the size of the
trusted or kernel infrastructure by creating ways to delegate out to a
service provided by untrusted and potentially enormous code. The
enthusiasts (such as Miller and Drexler) think of these potentially
enormous swarms as actually enormous and long-lasting; I think it's fairly
likely that one or a few strategies per market eventually mostly defeat
everyone else, so that the system as a whole is not particularly large.

However, the point is that you don't necessarily have to choose between
LZ77 and LZW or whatever.

Johnicholas

--bcaec54fbaf66b75e904c52e865d
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

You give compression as an example of a powerful computation primitive.<br>=
<br>Are you interested in the similarity between Matt Mahoney&#39;s ZPAQ vi=
rtual machine<br><a href=3D"http://mattmahoney.net/dc/zpaq.html">http://mat=
tmahoney.net/dc/zpaq.html</a><br>
which implements predictive compression by mixing a swarm of predictors, ea=
ch expert at a different domain.<br>And Miller and Drexler&#39;s idea of &q=
uot;Agoric Computing&quot;<br><a href=3D"http://e-drexler.com/d/09/00/Agori=
csPapers/agoricpapers.html">http://e-drexler.com/d/09/00/AgoricsPapers/agor=
icpapers.html</a><br>
where they propose routinely combining untrusted agents into reliable servi=
ces.<br><br>I think your focus has been on eliminating the software complex=
ity caused by history.<br>But there&#39;s another slightly different focus =
- reducing the size of the trusted or kernel infrastructure by creating way=
s to delegate out to a service provided by untrusted and potentially enormo=
us code. The enthusiasts (such as Miller and Drexler) think of these potent=
ially enormous swarms as actually enormous and long-lasting; I think it&#39=
;s fairly likely that one or a few strategies per market eventually mostly =
defeat everyone else, so that the system as a whole is not particularly lar=
ge.<br>
<br>However, the point is that you don&#39;t necessarily have to choose bet=
ween LZ77 and LZW or whatever.<br><br>Johnicholas<br>

--bcaec54fbaf66b75e904c52e865d--

--===============0442470378==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

-- 
To unsubscribe: http://lists.canonical.org/mailman/listinfo/kragen-discuss
--===============0442470378==--