[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Another explanation for dain daiin...

At 10:47 21/01/02 +0000, Gabriel wrote:
On 21 Jan 2002 at 10:40, Nick Pelling wrote:
> but here's my latest prediction for what dain / daiin is.
> "d<...>" means "copy the whole word <...> words back".
> This is a form of data compression (to be more precise, it's the basis
> behind the popular "Lempel-Ziv 77" algorithm): the offset field is
> normally heavily biased towards low numbers (because locality of
> reference usually ==> locality of relevance).

If it was a compression mechanism, the text would no exhibit so low entropy. The low entropy is not given
by daiin only. There is a strong bias for certain duplets as Stolfi pointed out some time ago.

On the contrary, this actually reinforces the duplets observation. :-)

I believe that the underlying alphabet is (with a few notable exceptions, such as "d" and "y") entirely expressed as duplets. However, this alone would lead to an overly-verbose ciphertext, with word patterns that we would frequently be able to pick out within a page.

Having a mechanism to copy nearby words (as described) would function not only to help keep the size of the ciphertext down to a reasonable size (useful if you're copying from an existing work with an otherwise verbose cipher), but also remove obvious cipher attack targets within a page.

Cheers, .....Nick Pelling.....

PS: this also points to a local per-page cipher mechanism being in use here, as this would reduce the number of inter-page duplicate symbols independently. For that, I look to the first 8-10 characters following the start of a page (which frequently seem to have different statistics from the rest of the page, when looked at as duplets).