Hacker News Reader

>This strikes me as very weird, since there is a Tocharian language group and Iranic languages don't belong to it.

I guess it's because "Tocharian" as used for IE languages of city-states of Tarim considered to be a mistaken name now. Eteo-Tocharian means True/Real Tocharian, and doesn't mean it belongs to "Tocharian".

   By 1851, Hincks and Rawlinson could read 200 Akkadian signs. They were soon joined by two other decipherers: young German-born scholar Julius Oppert, and versatile British Orientalist William Henry Fox Talbot. In 1857, the four men met in London and took part in a famous experiment to test the accuracy of their decipherments. Edwin Norris, the secretary of the Royal Asiatic Society, gave each of them a copy of a recently discovered inscription from the reign of the Assyrian emperor Tiglath-Pileser I. A jury of experts was impaneled to examine the resulting translations and assess their accuracy. In all essential points, the translations produced by the four scholars were found to be in close agreement with one another. There were, of course, some slight discrepancies. The inexperienced Talbot had made a number of mistakes, and Oppert's translation contained a few doubtful passages which the jury politely ascribed to his unfamiliarity with the English language. But Hincks' and Rawlinson's versions corresponded remarkably closely in many respects. The jury declared itself satisfied, and the decipherment of Akkadian cuneiform was adjudged a fait accompli.

https://en.wikipedia.org/wiki/Decipherment_of_cuneiform

To be clear, that is a property of some cryptography; without knowing the key any decryption is equally likely (i.e. there's no one 'valid' decryption result)

I'm not sure what you're saying?

Say you use an encryption method to encipher the first page of A Tale of Two Cities. Someone else is tasked with determining what the plaintext was.

If their methodology is to (1) choose some method by which the ciphertext is transformed into a candidate plaintext, and then (2) apply that method, they will not be able to produce a candidate plaintext which (a) makes sense but (b) is not the actual original plaintext.

The exception is one-time-pad encryption, which isn't really worth noting. It's an exception because the decryption method for one-time-pad encryption is isomorphic to "choose what you want the plaintext to say, and then write down the key that will produce that plaintext". If the key is shorter than the ciphertext, this can't happen.

But "a property of all encryption methods other than OTP" is a less convenient way to say "a property of all encryption methods".

...and you get a stretch of valid language...

How do you know it's valid?

This is actually a good question. Before the Rosetta stone work there were a number of published "translations" of Egyptian hieroglyph texts that were honest attempts at decoding the written language (as opposed to frauds). Typically they were somewhat nonsensical and this fact was papered over with "it's a religious text". In my opinion we see the same thing with proposed translations of yet undeciphered scripts like Rongorongo. One thing these typically share is they rely heavily on proposed ideograms and have little or no phonetic component, because there's more room to massage meaning with ideograms.

>> However, ancient languages are usually not understood that well, which makes it impossible to determine whether a decoding really is or isn't valid.

... you can read and comprehend it?