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

RE: VMs: Tironian notes

Hi GC,

At 19:00 29/04/2003 -0500, GC wrote: 
Nick wrote:
> I believe that the structural difference between
> occurrences of "4" and
> "4o" is a key indicator of what is going on "under the
> hood" of the VMS'
> coding system. To my eyes, they function as autonomous
> letters - in the
> same way that "o8" and "89" function as autonomous
> letters, for example -
> which is why I've moved so strongly towards reading the
> bulk of the text in
> pairs.

The VMS glyphs don't have to represent numbers in order to operate
in pairs.  Vig and Selenus cover several double letter ciphers
that use 'standard' alphabets.

I absolutely agree, that was just a coincidence in the choice of pairs I gave. I'd be pretty sure that <o + gallows>, <y + gallows>, <ol>, <or>, <os>, <ar>, <chy>, <shy>, <cho>, <sho>, <che> (to name but a few) are almost certainly pairs coding for a single underlying token, at least some of the time. :-)

I guess my problem here is that reading the VMS symbols in pairs
effectively reduces the transmittable information by half, does it
not?  The 'words' are "relatively" close to Latin in their present
form, but if it takes two glyphs to make an information bit,
wouldn't the words have to be about twice as long as the
underlying language?

The (perhaps surprising) answer is "not necessarily". For example, imagine constructing a list of syllables (ba be bi bo bu, etc), do a frequency analysis on them (for your text), and then take (say) the most common 50 - and let's say these comprise (say) 70% of the text's contents (finding out an actual value for this for various languages circa 1500 would be an excellent statistical exercise to carry out, but one which I have shamefully failed to do to date).

If you then assign a cipher pair for each of these common syllables, and a cipher pair for each of the remaining letters, the approximate length of the text would be:-

        70% x 2 / 2     = 70%
        30% x 2 = 60%
        Total           = 130% of original size

...rather less than 200%. It may well be that if you condense other consonant pairs (such as "ch", which was often written as a single sign anyway circa 1500, and "pr"), the typical figure may come down closer to 100%... yet you completely control the apparent (glyph level) statistics!

The very first page of MS Vienna 2398 is the Tranchedino cipher, which Milan used to get messages to/from Florence - the central axis of Northern Italian diplomatic communication 1450-1500. To my eyes, this cipher shows clear signs of having been updated and extended numerous times (it's dated 1450, but I believe that's only when the original core of the cipher was added to the ledger) - the page is littered with symbols for syllabic and consonantal groups, very much in the kind of way I'm describing.

You can have all sorts of theories about artificial word division
and so forth, until it comes to pages with labels.  The labels are
approximately the expected necessary length to communicate
information, and halving this would render the labels
'ineffective'as information carriers???
The one on the top right of f89r1 for instance, is about the right
length for "serpentia", which would seem to be indicated by the
snake root in the drawing.  Cut this in half and you can barely
fit in "snake"!  It's little things like this that add up and
leave me with the distinct impression that each glyph is a
transmission unit.  Any thoughts?

Labels are definitely tricky, though (because of the preceding discussion) perhaps not ~quite~ for the length-related reasons you give.

If you look at everyone's favourite source of easy (but false) answers in the VMS - the Pisces diagram - I would argue that the likelihood of labels like "oty" containing full-blown word-like information are fairly small. Even taken pair-wise, the famous <otolal> in the centre (which, of course, would probably be composed of <ot-ol-al>) is fairly enigmatic - even if <ot> represents a letter (or even a syllable), why would so many labels begin with it?

I think the answer perhaps comes on the following pages: if you look at Aries and Taurus (which have been split into two volvelles each), the labels are nearly twice as long as on most of the single-sign volvelles (like Pisces) (look at the central ring of Sagittarius, for example!)

From this, I infer that the labels on the single-sign volvelles are compressed - perhaps abbreviated, or perhaps number indices into a separate table (perhaps in the rings of text or elsewhere, who knows?), and that the labels on the light-dark split-sign volvelles are probably closer to real text.

For that reason, if I was asked to point to the VMS text likely to be closest to plaintext (or, rather, the "least heavily obscured text"), I'd suggest the labels on Aries and Taurus. Sadly, I don't know for sure what the plaintext would be - but would be unsurprised if they were (as I've written before) numbered sequentially clockwise from 10 o'clock (the ascendant, where the ornate little boxes are on the text rings on the dark Aries volvelle), inner-ring outwards, and named the rising stars for each degree in a tradition following Pietro d'Abano. Just a rough guess, you understand. :-)

Finally: as far as f89r1 (and in fact all the antidotarium) goes, I'd guess that the labels are most likely simply to be numbers and/or apothecary shorthand. Your "serpentaria" label ends with <daiin>, which I guess is "oncia tria", three ounces. As for <ar-ar-cho->, I don't know - but to my eyes, <ar-ar> seems too well-structured to be text or cipher, except if considered as pairs.

I certainly suspect that, because they occasionally appear repeated in sequences (and I've flagged on instance where the author appears to have tried to hide this), <ar> and <or> are most likely to code for "M", "C", or "X". If I was making a code 1400-1500, I'd have been sure to discard repeated symbols wherever possible... but one place where you wouldn't really have that latitude would be in Roman numerals. And there's no evidence of any other numbers! :-o

Anyway, I guess that's just about enough unsupported inferences for one day. :-)

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


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