Re: [colorforth] shifted huffman trees

[Mark Slicker <maslicke@xxxxxxxxxxx>]

On Fri, 24 Sep 2004, Bill Parker wrote:

> --- Mark Slicker <maslicke@xxxxxxxxxxx> wrote:
>
> > For the bit-stream you would need a prefix-code
> > because the tags are variable length. For a known
> > word-length, you are right, that should work.
>
> I realize now that my past posts (which all tend to
> just talk about 'encoding') have not been at all clear
> that I was only talking about the encoding of
> information in the target machine's memory.  I think
> that's part of the reason I had trouble understanding
> some of Mr. Moore's criticisms.
>
> I haven't given much thought to transmitting
> information in a bit-stream because I don't have two
> machines running colorForth.  But I would imagine it
> would be very different than how one might choose to
> store the preparsed data on the target machine.
>
> For example, I'm not even sure I would send tags at
> all.  The bit stream might look a lot more like
> operator input.  I could see keeping a 'space'
> character to break between words of the same type
> (tag) and only sending tag-like information when the
> type of word changes.

My colorForth to html converter does this, however it is intended as a
one way translation. With prefix-coding, I think you would need a special
code to change colors, or else use run-length coding which Chuck
has stated as a possibility.

>  Numbers would probably be sent
> variable-length with the most significant bit sign
> extended on the receiving end.  (I think someone here
> already talked about that).
>

I wrote this compressor which uses variable-length numbers:

http://personalwebs.oakland.edu/~maslicke/colorforth/compress/

---------------------------------------------------------------------
To unsubscribe, e-mail: colorforth-unsubscribe@xxxxxxxxxxxxxxxxxx
For additional commands, e-mail: colorforth-help@xxxxxxxxxxxxxxxxxx
Main web page - http://www.colorforth.com