Conversation
hot take
Show content

i don't think cryptographers should be allowed to write cryptographic specifications

i understand that this is very precise but if you're trying to implement the algorithm it adds _so_ much overhead

2
0
0
re: hot take
Show content

this is an xor operation described as "constant addition", right

2
0
0
re: hot take
Show content

is the and operation implicit

what is the operation precedence here

edit: okay, python actually has the right xor precedence

2
0
0
re: hot take
Show content

@whitequark Reminds me of PL research with its arcane notation involving a lot of Γ, ∆, ⊢, and τ.

Also reminds me of this video: https://youtu.be/33y9FMIvcWY?si=FjgQDjaKryIvyQUu

2
0
0
re: hot take
Show content

@lesley it took me several years to grasp the PL research notation, but at least there it adds actual value

the Ascon spec feels like it's written to be obscure. the algorithm is very simple, it's just described in a crackhead manner

0
0
0
re: hot take
Show content

what would possess someone to describe xor with a 8-bit number as a 64-bit addition

4
0
0
re: hot take
Show content

@whitequark you have the context, we don't, but why wouldn't “⊕” represent “addition modulo 2n”, or “addition” for short, here?

1
0
0
re: hot take
Show content

@void_friend because this is a part of a broad family of ciphers which also use addition with carry (i'm upgrading from blake3/chacha20 to something that occupies less of the fpga)

0
0
0
re: hot take
Show content

@whitequark carryless addition /j

1
0
0
re: hot take
Show content

@charlotte i understand this, i just think this is confusing given that carryful addition is _also_ used in similar algorithms

0
0
1
re: hot take
Show content

@whitequark Is this because to them it is natural to talk of (XOR, AND) as a ring and thus to use ring notations?

1
0
0
re: hot take
Show content

@MonniauxD think so

0
0
0
re: hot take
Show content

this is going to be implemented in a programming language. copied from a pdf and symbols replaced

omitting multiplication operators, while natural from a math point of view, makes this _so_ much more painful

1
0
0
re: hot take
Show content

@whitequark honestly seems more like the kind of thing that would end up in some "is this primitive turing complete" ML paper

0
0
0
re: hot take
Show content

@lesley @whitequark god this video is so so bad. guy who does not know any prerequisites for the topic he wants to learn about expects wikipedia to be a textbook

1
0
0
re: hot take
Show content

@uniwuni @lesley @whitequark and so, wikipedia continued to be incredibly, notoriously and discouragingly useless for mathematical topics while being approachable by non-experts in every other subject…

0
0
0
re: hot take
Show content

cryptographers: addition is +, xor is ⊕

same cryptographers: now we're gonna use the word "addition" for xor

it's not that i'm missing something, this paper is just poorly written

1
0
0
re: hot take
Show content

also they don't specify the endianness of concatenation. why is it pointlessly verbose where it doesn't matter, and not verbose enough where it does

0
0
0
re: hot take, unicode, unhelpful
Show content

@whitequark did you copy-paste the text (or otherwise inspect it) to see if there is a INVISIBLE TIMES (U+2062) in between?

1
0
0
re: hot take, unicode, unhelpful
Show content

@timotimo *stare*

1
0
0