[TML] New 3d computer simulation design

[Patrik Holmstrom]

On Dec 3, 2007 6:54 AM,  <shadow at shadowgard.com> wrote:
>
> On 3 Dec 2007 at 5:01, Patrik Holmstrom wrote:
>
> > On Dec 3, 2007 4:16 AM,  <shadow at shadowgard.com> wrote:
> > > Somewhere in the archives is a discussion of formats and cost for X-
> > > boat messages. At a guess, 10 years back.
> > >
> > > We had it broken down using then current stuff and a bit of guessing
> > > as to future tech.
> > >
> > > For example, based on per bit or byte charges, it's noticeably
> > > cheaper to use a limited character set (Baudot or ASCII) than to use
> > > something like unicode.
> >
> > Yeah, but the Shannon entropy of the data is still the same so if you
> > compress it you end up with identical (theoretical case) or near
> > identical (real world case) sizes. For instance Mary Shelly's
> > Frankenstein goes from (ascii/unicode) 439kB/877kB to 140kB/142kB
> > using the 7z compression algorithm. Since you'd like to compress the
> > message anyway you don't save anything useful from using a more
> > compact character set.
>
> On the other hand, compressing the data makes it more vulnerable to
> errors in transmission.

If the channel is error prone it's going to have some sort of error
correction. Compressed data (and that would also include images,
sound, 2D/3D video) is not the most vulnerable data, consider source
code/compiled code, finacial data, encrypted files, et.c.

> BTW, have you tried seeing what size Frankenstein would be in Baudot?

A quick look around couldn't find any converters (although I found a
lot of software teletype emulators) but uncompressed it should be
5/8th the size of ASCII and compressed it would be the same size as
compressed ASCII or a hair smaller. The original encoding is of little
significance since compression replaces it by (in theory) the typical
set or (in practice) Huffman/Lempel-Ziv/etc encoding. A good
compression algorithm is really a pretty accurate estimate of the
amount of information in a piece of data. Try for instance to compress
something like PI in binary form (a patternless sequence of binary
digits with no wasted encoding).

/Patrik