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

Re: VMs: Pairing



Hi Barbara,

I really like your idea of paired letters being significant (and I see
why spaces wouldn't be word division in such a case), after all without
using code words or syllabics it's difficult to shrink words but easy to
expand them. However a number of things puzzle me if that notion is
valid.

The first is that if every pair of letters is significant (effectively a
single grapheme) it would half the length of the Mss "letter" count, and
thus also half its potential information content.

If you combine this with shorthand (discarding the roughly 50% of the original letters which are redundant in typical texts), then the net change in size would be nil. It's not ~quite~ that simple here, but that's the basic principle I think is at play here - give with one hand and take away with the other. :-)


The second is where does this leave "stand alone" glyphs that are
between "words"? These have spaces on either side of them, one of those
spaces would have to be a "division" of some sort wouldn't they?

The first thing to note is that a "pure" pair cipher would be fairly useless, for the reasons Jacques listed - first-order context effects would re-appear (though somewhat diluted) as second-order effects, second-order as fourth-order, and so on. I'm not proposing that the VMS is a pure pair cipher! :-)


What I am proposing is that the VMS' cipher appears to be a sophisticated pair cipher, which uses a tricky combination of single letters and paired letters. For example, in my pair-transcription scheme, EVA <s> is a single letter, but <r> always appears as part of a pair/group (<or>, <ar>, <air>, <aiir> etc). It's true that there are a handful of <os> pairs in the text (18, IIRC), but these are so very few (as compared to <or> pairs) that I'm happy to put them down to either copying or transcription errors (there's an issue about <o> in combination with <ch>/<sh>/<e>, but that's another story).

Another standalone example is that I transcribe <o[gallows]> and <[gallows]> as separate letters: and whereas <l> appears very nearly all the time as part of <ol> / <al>, it also sometimes appears as a standalone (ie unpaired) character (typically in the balneological section).

So, I have no problem at all with standalone characters, but (as you point out) spaces may or may not be indicative of anything. Until I have evidence either way, I prefer to ignore spaces - and I don't know of any coding system from this general period that encoded non-space-like information in spaces (there's actually a modern programming language which is written entirely in whitespace, but that's another story again).

The third is that if the paired cipher idea is valid then the total
number of glyphs would be an even number would it not? If the number is
odd then there is at least one glyph with an independent value. And if
there is at least one glyph with a unique value, why not others? That
being the case I don't understand how one could determine which pairings
are valid and which are not, which glyph(s) are singly significant and
which are not. So, how will you go about making such a determination?

Through trial and error, I've been iterating towards a set of rules for deducing the pair locations within EVA text. I posted a Bitrans script recently that should get most of the way there: the remaining complexity is in the {e|ee|eee|eo|ech|eche|eeeo|...} combinational group, which (strictly speaking) I think is another cipher entirely (that I believe codes for word endings).


The fourth is how many pairings actually (or potentially via different
counts of the Mss) actually exist? The hypothetical number of possible
pairings must be huge (the heat here in southern england is frying my
brain so I can't recall if formula's n^n or [n!] ). Even allowing for
contractions, abbreviations, truncations etc etc surely only a few
hundred would be needed to encipher any language thus. If it turns out
that the pairings in the VMS run into several hundreds, or even
thousands, where does that leave the pairing theory?

In practice, there aren't really that many - this list pretty much covers all of them (apart from the {e|ee|...} group aside, as mentioned before):-
<or> <ar> <s> <ol> <al> <am> <qo>
<[gallows]> <o[gallows]>
<c[gallows]h> <oc[gallows]h>
<d> <y> <dy> <od> <l>
<ai[i][i]n>
<ai[i][i]r>
<ch>/<sh> <-- I predict these are truncation tokens


That's rather less than "thousands" (in fact, it's about 35 or less) - try it for yourself! :-)

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


______________________________________________________________________ To unsubscribe, send mail to majordomo@xxxxxxxxxxx with a body saying: unsubscribe vms-list