The Nineteenth Byte

DJMcMayhem

22:00

It's weird that you're helping the robbers so much

You don't need to do that

seshoumara

@Mego Has the counting so far been problematic to decide who won a golf challenge? In every special case I think the actual count would be higher or equal in encoded bytes compared to its own count.

Anonymous

@seshoumara It hasn't been problematic in that way, but it can be confusing at times when trying to determine whether or not a given scoring method is acceptable. It violates the principle of least surprise.

Anonymous

And the fact that one scoring unit may encode a different amount of information than one byte makes it potentially unfair

seshoumara

Like I said, in those special answers that is mostly always the case.

And what about those awesome, out of the box solutions like the ones using real-life objects, like pulleys, weights, dominos, chess pieces etc.? Would these become invalid answers in your new system?

Anonymous

@seshoumara They're already invalid because they're not reproduceable

seshoumara

22:06

not all of them

Anonymous

14

A: Are physical analogs of programming legitimate?

To be acceptable as a programming language, a digital simulator must exist. Photos of dominos are great fun and all, but by themselves they don't really meet the definition of programming language. The basic problem with programming languages 'implemented by the physical world' is that the resul...

seshoumara

the logical gates done by pulles and weights are for ex

Anonymous

> To be acceptable as a programming language, a digital simulator must exist.

seshoumara

sure, that's the person's hand doing the atomic actions: hold pulley immobile, let this pulley be mobile, etc

a, digital only. pff

Anonymous

You may disagree with the policy, but community consensus has decided that physical analogs are not allowed unless there is a digital simulator that can reliably reproduce the "program"

seshoumara

22:11

which in most of those creative cases, there can't be a digital simulator. Hmm, but would a chess engine that applies chess moves from a given position be such a simulator if I somehow find a way to calculate sth with chess pieces?

Mistah Figgins

It's important for the "program" to be reliably reproducible, ideally at little cost to the tester (in time and effort). This means, 99% (if not 100%) of the time, a computer program. If you have to write it, so be it.

seshoumara

interesting

ais523

I'm still upset at the requirement that a program must be a whole number of bytes long

and it's currently unclear what circumstances could change that requirement

(not all computers even use 8-bit bytes, although basically all computers used nowadays do)

Roujo

Wait, so a custom program spec that handles sub-8-bit bytes wouldn't be allowed?

ais523

I wrote the 7 interpreter so that it would work regardless of the number of bits in a byte

Anonymous

22:20

@ais523 You would need a computer that supported bit-level addressing and a storage system that allowed storing and fetching individual bits

ais523

and yet people still seem to round the 7 programs up to a whole byte

@Mego I've been considering writing a filesystem for the purpose

however I think I'd make it also support fractional numbers of bits

Anonymous

@Roujo It would be allowed, but you can't have (for example) 24.25 bytes because no computers nowadays are capable of fractional bytes. It would be stored in memory as 25 bytes.

ais523

that way you could store a lot of information in the length of the program

Roujo

@ais523 Fractional bits might be overkill :P

Anonymous

@ais523 "fractional numbers of bits" doesn't make any sense

ais523

22:22

@Mego I mean, that's not even true when you're talking about the CPU itself; x86's addition instruction adds together two 64-bit numbers into a 65-bit result

Anonymous

A bit is the fundamental unit of information. You can't have 0 < n < 1 bits.

Roujo

@Mego Don't we allow assembly languages for architectures that don't really match modern computers already?

ais523

@Mego you can have 1 < n < 2 bits, though

Anonymous

@ais523 Sure, but that one extra bit is stored in a register (whose other bits are used for other purposes)

ais523

it's the amount of information in a trit

Anonymous

22:23

Oh, so you're talking about allowing other systems than binary. Gotcha.

Jan Dvorak

Then you can have <1 bits as well

ais523

and something can also have 0 < n < 1 bits in a compressed representation (this basically means that appending it to a compressed string doesn't always cost a bit)

Jan Dvorak

in the Manchester encoding, each symbol represents half a bit

Roujo

@JanDvorak TIL, thanks ^^

Anonymous

@ais523 That's kind of a cop-out. You're looking at the uncompressed size of the appended part and the compressed size of the current part.

ais523

22:25

@Mego you're looking at the compressed sizes of both

you can say "this symbol costs less than a bit when compressed"

Jan Dvorak

Can I coin .rb.gz as a programming language?

ais523

@JanDvorak you mean the interpreter just unzips the input and feeds it to Ruby? yes, but I'd recommend something terser, such as raw DEFLATE without headers

Anonymous

@ais523 The compressed size of the appended part wrt the compressed size of the current part. On its own, no symbol can represent less than a bit of information.

ais523

zip has too much overhead to save bytes when you have short programs

@Mego if a symbol appears more than half the time, its appearance represents less than a bit of information

Jan Dvorak

Golfing ruby-deflate (or rather optimizing the ruby code so that its "compilation" results in the smallest deflate file) is an interesting challenge, though.

ais523

22:29

e.g. in an encoding where 0 encodes 00, 10 encodes 01, and 11 encodes 1, a 0 in the input takes up half a bit of the output (but a 1 costs two bits, to compensate)

Riker

@Roujo hey haven't seen you around in a while

Anonymous

I think we're having two different conversations

ais523

I think the problem is that one of us is conflating symbols, the encoding of symbols, and the information content of symbols; I think it's you but it might be me

Roujo

@Riker Life caught up, as it often does ^^

Riker

@Roujo ah, okay

welcome back!

ais523

22:41

hmm, now I'm busy reading documentation on writing filesystems

I can see how you can have a fractional number of bytes in a file

can you have a fractional number of files in a directory, though? I suspect not

(although I assume the file/names/ could reasonably have fractional size)

hmm, and now I'm wondering if a filesystem can be Turing-complete by itself

Jan Dvorak

What would half a file be?

Roujo

@ais523 I guess you could make up a filesystem where saying "The rest of the file is at /some/other/path", kind of like a linked list of sorts, and then say that only 1/nth of the file is in this directory.

ais523

hmm, right, you split the file up between directories

with half the filename and half the inode number stored in each

Roujo

But I can't think of a way in which that would be a better idea than how we currently handle, say, multipart zips

ais523

wait, no, full filename in each, half the inode number in each

it's not intended to be a good idea

Roujo

22:46

I guess the combining could be done at the FS driver level

Instead of the application level

Oh okay :P

ais523

this is purely about creating corner cases and uncovered situations in PPCG rules; I like esoteric programming, which in many ways is all about finding assumptions to break, and the PPCG rules are a nice shiny list of assumptions standing there ready to be broken

Roujo

Kind of like you can have an NTFS volume span multiple physical disks, although I've never heard someone count volumes as fractions ¯\_(ツ)_/¯

Jan Dvorak

I don't think a filesystem can test primality...

Roujo

@ais523 Yeah, I know you enough to have figured that's the case :P

ais523

@JanDvorak well you could do something trivial like making symlinks fail to resolve if they pointed at files whose name was a composite number, but that'd be really arbitrary

hmm, what about supporting regex symlinks? with a sufficiently powerful regex syntax that's Turing-complete

Jan Dvorak

22:51

@ais523 esolangs.org/wiki/Bubblegum

ais523

yes, I know of Bubblegum

Roujo

I was thinking of how a filesystem could compute anything at all without a driver, but then again it's just a spec with an entry point. So input could be the FS structure, the entry point could be "access a file named /main" and the FS spec handles what happens then. Wow.

Destructible Watermelon

23:17

@JanDvorak stream is coming back

Riker

huh, playing portal 2 and asking only 2 questions about it on arqade gave me 2 popular question and 1 notable question badge

i.e. 3.5k distinct IPs looked at my questions o_O

quartata

23:30

Not distinct.

Does anyone know how efficient pthread_kill is?

Riker

@quartata views != distinct? TIL

quartata

Like in terms of how long it takes for the other thread to receive the signal

Riker

oh cool, meta.se says it's any views, duplicate views from the same user/IP are only recounted 15 minutes after the last one

Destructible Watermelon

hmmm, I seem to not have the latest DF version

Riker

please don't start redirecting idle DF chatter into here, people have already expressed their displeasure with nethack talk

Destructible Watermelon

23:44

ok

Transcript for