"Tim Tyler"  wrote in message
news:c9b74i$115$1{at}darwin.ediacara.org...
> Jim Menegay  wrote or quoted:
> > Tim Tyler  wrote in message
news:...
> > > Jim Menegay  wrote or quoted:
>
> [snip agreement]
>
> > > > So far, analog and digital seem identical - at least in
> > > > principle.  But when you start looking at the details, there
> > > > are differences.
> > > >
> > > > The error rate for a good digital operation is something
> > > > like one error per thousand.  That is 0.1% of the time,
> > > > the information is changed, whereas 99.9% of the time, it
> > > > is unchanged.
> > > >
> > > > The error rate for a good analog operation is something
> > > > like one part per thousand.  That is, 100% of the time the
> > > > information is changed, but it is not changed by much -
> > > > only by about 0.1%
> > >
> > > I am not aware of anything about the definition of the
> > > terms "analog" and "digital" that
specifies what error
> > > rates they exhibit.
> >
> > Perhaps I was unclear.  My numbers were intended as an
> > illustration that both analog and digital are subject to
> > noise, but that the noise manifests itself in different
> > ways.  I did not mean to suggest the particular error
> > rates that I used in my example as part of the definition
> > of digital or analog.
> >
> > > In the real world, both sorts of system are subject to
> > > noise - and neither has perfect fideleity.  If you assume
> > > that analog storage media have high error rates, then
> > > your conclusion follows.
> >
> > I made no such assumption.  I tried to make the error
> > rates equivalent.
>
> A 1 chance of a bitflip every 1000 bits is roughly equivalent
> (in terms of the error rate) to a +/- 1/1000 change in an 0-999
> integer value?
>
> It isn't.  It isn't anywhere remotely equivalent:
>
> Store 1000 bits of information in both media.
>
> That's 1000 bits of digital information - and
> about 98 numbers from 0-999.
>
> How much information is needed to store the details of
> what errors took place in one generation?
>
> In the digital case a number between 0 and 999 should
> normally suffice to describe the location of the error.
> Sometimes more will be needed - but sometimes less -
> so the error can be described in 10 bits.
>
> In the analog case you need about one bit per value to
> encode whether the error is "+" or "-".  So that's
about 98 bits.
>
> About ten times much information is needed to describe
> the locations of the errors in the analog case.
>
> That's about the same as saying there was ten times as much error
> occurring in the analog medium.
>
> No wonder it doesn't make such a good information storage device,
> in your example - the error rate is effectively ten times as big.
>
> This doesn't reflect badly on analog media, though - it just
> shows that the error rates you pulled out of the air happened
> to heavily favour digital media.

Hmmm.  You seem to be correct, and my intuition was incorrect as
to what would be equilivalent error rates between analog and
digital.

However, since I did not use these error rates in my argument,
I think that my argument still stands.  And the argument was
that analog media, whatever their benefits for short term
use, have severe problems when compared to digital as a medium
of biological heredity.  Those problems stem from the fact that
the two media have qualitatively different failure modes in the
presence of noise.  Analog data tends to drift, whereas digital
data tends to break.  My intuition, which is now proved to be
suspect ;-(, is that digital is better because it permits
more effective proofreading.  With digital, you are proofreading
against a standard that is possibly broken, but probably not.
With analog, you are proofreading against a standard which
has certainly drifted.  My intuition says that digital
proofreading works better.  But to turn this into a proof,
I am going to have to understand better what constitutes an
equivalent error rate between the two media.  Any links that
you can provide related to information theory would be
appreciated.
---
þ RIMEGate(tm)/RGXPost V1.14 at BBSWORLD * Info{at}bbsworld.com

---
 * RIMEGate(tm)V10.2áÿ* RelayNet(tm) NNTP Gateway * MoonDog BBS
 * RgateImp.MoonDog.BBS at 5/31/04 5:02:45 PM