156829? Mass-market. Tacky.
February 4, 2016 4:34 AM   Subscribe

Artisanal Integers - Summer of 2012. Suddenly several “integer-as-a-service-providers” spring from nowhere. They deliver “artisanal integers”. Integers which (they claim) are “hand-crafted and guaranteed to be unique and hella-beautiful”.
posted by Wolfdog (78 comments total) 14 users marked this as a favorite
 
689140758
posted by marienbad at 4:50 AM on February 4, 2016


Whole foods, whole numbers.
posted by Segundus at 4:54 AM on February 4, 2016 [5 favorites]


8675309 was auctioned off to an anonymous buyer known only as "Jenny."
posted by Faint of Butt at 4:57 AM on February 4, 2016 [34 favorites]


689140758

I'm sorry marienbad, that's not Numberwang.
posted by sektah at 5:10 AM on February 4, 2016 [49 favorites]


They're all so ... small ...
posted by mr vino at 5:18 AM on February 4, 2016 [3 favorites]


23 Integers Scientists Don't Want You To Know About (#57361 Will Blow Your Mind!)
posted by Mr.Encyclopedia at 5:20 AM on February 4, 2016 [26 favorites]


And I can't help but notice how much these "hipster" integers tend toward the positive side of the spectrum...
posted by Navelgazer at 5:21 AM on February 4, 2016 [1 favorite]


Mapping needed between these services and this one.
posted by oheso at 5:23 AM on February 4, 2016 [2 favorites]


I sense a market opportunity. Artisanal hexadecimal numbers. FEB04!
posted by tallmiddleagedgeek at 5:26 AM on February 4, 2016 [7 favorites]


Um...Can someone assure me this is merely a joke I'm not understanding? It's hard to tell sometimes.
posted by Thorzdad at 5:26 AM on February 4, 2016 [3 favorites]


Whole foods, whole numbers.
Add numbers. Not too many. Mostly integers.
posted by thelonius at 5:27 AM on February 4, 2016 [27 favorites]


It's something serious wrapped up in a joke, while still also being serious. It's taking a legit technical need and throwing a wink and a nod on top.
posted by Itaxpica at 5:31 AM on February 4, 2016 [6 favorites]


This sort of closed minded experience is exactly the reason I've named my latest work 34723901 + .1
posted by eriko at 5:33 AM on February 4, 2016 [3 favorites]


Artisinal? Sure. Quality? Probably. Prime? Not so much.
posted by penguinicity at 5:42 AM on February 4, 2016 [9 favorites]


* Prime as in prime rib.
posted by mhoye at 5:44 AM on February 4, 2016


leave it to silicon valley to reinvent the phone book.
posted by indubitable at 5:45 AM on February 4, 2016


It's taking a legit technical need

No, it's just plain stupid.
posted by effbot at 5:47 AM on February 4, 2016


2305843008139952128
posted by sammyo at 5:47 AM on February 4, 2016


Well, sure they're natural numbers, but are they also organic?
posted by Mayor West at 5:51 AM on February 4, 2016 [6 favorites]


If there is a legit technical need, can someone please explain?
posted by trif at 5:51 AM on February 4, 2016 [1 favorite]


They're trying to avoid half measures.
posted by vbfg at 5:53 AM on February 4, 2016


the RAND Corporation actually used to publish an atlas of random number values; I think it may be online somewhere
posted by thelonius at 5:58 AM on February 4, 2016 [4 favorites]


But.... but... they gave me a fraction.

Don't know if I should feel insulted or special.
posted by RolandOfEld at 6:01 AM on February 4, 2016


I just want a small set of irrational reals other than e, pi and phi. Is that so hard?
posted by Octaviuz at 6:07 AM on February 4, 2016 [2 favorites]


The About page at London integers does explain a bit more about the serious idea underlying the jokey "artisan" stuff.
posted by Wolfdog at 6:07 AM on February 4, 2016 [3 favorites]


I use Gaussian integers, but you probably wouldn't have heard of those.
posted by escabeche at 6:11 AM on February 4, 2016 [17 favorites]


If there is a legit technical need, can someone please explain?

what they want is some way to uniquely label things across more than one program.

if you have a single program, then you can just call the first thing you have "1", the second "2" and so on.

if you have multiple programs, then they need to agree amongst themselves so that they avoid using the same number in two places. or, as here, use a separate third party source of numbers (so one place is responsible for making sure each number is used just once, and each program goes there to get a number when they need one).

BUT there are other solutions that make more sense than this (the things called UUIDs in the article). so this isn't really necessary. the only advantage of this approach is that the numbers can be smaller, which saves a little disk / memory space. but really that's not a big concern in almost all cases.
posted by andrewcooke at 6:11 AM on February 4, 2016 [3 favorites]


so, basically they are trying to use something diffirent instead of UUID scheme because the UUID scheme provides strings of alphanumerical characters and dashes which are (supposedly), less efficient at the database layer and these guys use only integers. And their (these guys) way to prevent duplication is based on an honor system.

Full marks for fun, but 0 for sensibility, scalability and utility.
posted by TheLittlePrince at 6:13 AM on February 4, 2016 [1 favorite]


marienbad: "689140758"

You didn't tell it right!
posted by chavenet at 6:22 AM on February 4, 2016 [5 favorites]


so, basically they are trying to use something diffirent instead of UUID scheme because the UUID scheme provides strings of alphanumerical characters and dashes

UUIDs are 128-bit numbers. Their most common representation is with hex digits and dashes, but at their root they're just 128-bit numbers.
posted by Slothrup at 6:23 AM on February 4, 2016 [6 favorites]


Yes, exactly. This whole thing makes zero sense (zero being, incidentally, an integer).
posted by a car full of lions at 6:28 AM on February 4, 2016 [1 favorite]


And I can't help but notice how much these "hipster" integers tend toward the positive side of the spectrum

That's because they're all natural numbers!
posted by Obscure Reference at 6:40 AM on February 4, 2016 [3 favorites]


Octaviuz I just want a small set of irrational reals other than e, pi and phi. Is that so hard?


Not hard at all. We can choose some irrational algebraic numbers. For example: Any n-root of any prime is irrational (p^(1/n) where p is prime and n is natural).

Now if you want a set of transcendental reals, that's harder. There's a lot more trancendentals that algebraics, but specific trancendentals tend to be much harder to describe.
posted by yeolcoatl at 6:47 AM on February 4, 2016 [5 favorites]


I mean I suppose you generate trancendentals with something like z*pi (where z is an integer), or a*e (where a is algebraic), etc. But that's not really satisfying.
posted by yeolcoatl at 6:49 AM on February 4, 2016 [3 favorites]


So, UUIDs are statistically unique. You call a function that generates a random value, and the result is highly, highly unlikely to match any other UUID generated by any other call to any other function that also generates UUIDs. But, there's no way to actually guarantee uniqueness unless you actually have a centralized service that says, "No, don't use that one, it's taken."

More importantly, though, UUID generation on a local machine relies on the weak entropy generators that most computers ship with. Usually not a problem, but sometimes you need a better source of entropy that most computers have, and good entropy generators are really expensive.

Anyway, I can't tell you what the precise motivation for any given integer-as-a-service site is, but I can think of at least one application that needs good entropy, uniqueness, and integers rather than UUIDs. Namely, cryptographic salts.
posted by tobascodagama at 6:49 AM on February 4, 2016


escabeche: I use Gaussian integers, but you probably wouldn't have heard of those.

I can't imagine what you're talking about.
posted by vernondalhart at 6:51 AM on February 4, 2016 [5 favorites]


I'm looking to buy a house in the Diophantine District, do you have any leads?
posted by oceanjesse at 6:56 AM on February 4, 2016


Where is Ramanujan when we need him?
posted by ubiquity at 6:58 AM on February 4, 2016 [1 favorite]


+1
posted by Devonian at 7:06 AM on February 4, 2016 [5 favorites]


I bet each one comes with pretentious care instructions about never storing them as strings.
posted by ignignokt at 7:06 AM on February 4, 2016 [16 favorites]


So, UUIDs are statistically unique.

there are various kinds, including one that uses mac address and time. so as long as you are confident mac addresses are unique, and you don't exceed the rate for unique times, those are "guaranteed" unique.

also, there comes a point where "statistically unique" is as close to unique as "really unique + random errors" (i imagine it's around 256 bits, not 128, but that's just a wild guess)

cryptographic salts

these usually don't have to be globally unique - only unique to any particular private key - so you can use a local counter.
posted by andrewcooke at 7:25 AM on February 4, 2016 [1 favorite]


Where is Ramanujan when we need him?

I think he lives at 1729 Hardy St.

(And speaking of the importance of picking the right numbers... Network tool contained hard-coded prime number that wasn't prime after all.)
posted by kmz at 7:27 AM on February 4, 2016 [4 favorites]


UUIDs may be statistically unique, but they are not authentically unique. They lack the artisanal quality that make the Brooklyn, Mission, and London integers so special. Anyone can generate a UUID and expect it to be unique. Might as well call those "plebian integers".
posted by Nelson at 7:33 AM on February 4, 2016 [7 favorites]


No mention of Canal St. Integers, typical.

(I made Mission Integers, and Nelson Minar is Mefi’s own)
posted by migurski at 7:34 AM on February 4, 2016 [4 favorites]


the RAND Corporation actually used to publish an atlas of random number values; I think it may be online somewhere

Here it is. The ebook is free, or you can buy a fine quality bound volume.
posted by thelonius at 7:37 AM on February 4, 2016 [1 favorite]


Network tool contained hard-coded prime number that wasn't prime after all

Isn't there a gag name for this? Based on a mathemetician who gave a non-prime small integer as an example of a prime, in a talk or article?
posted by thelonius at 7:38 AM on February 4, 2016


Grothendieck Prime
posted by Elementary Penguin at 7:46 AM on February 4, 2016 [6 favorites]


57 is sometimes called the "Grothendieck prime" for that reason. Curses!
posted by Wolfdog at 7:46 AM on February 4, 2016 [2 favorites]


Article also doesn’t cover what an incredible journey the integer space has been.
posted by migurski at 7:49 AM on February 4, 2016 [5 favorites]


Wasn't there a thing where, when everyone was using the RAND book of random numbers, there was a bias toward the numbers on the first few pages?
posted by rockindata at 7:49 AM on February 4, 2016 [2 favorites]


8008135
posted by JohnFromGR at 7:51 AM on February 4, 2016 [1 favorite]


121242121242484242121242121
posted by Spathe Cadet at 7:53 AM on February 4, 2016


Wasn't there a thing where, when everyone was using the RAND book of random numbers, there was a bias toward the numbers on the first few pages?

To avoid this risk, you can read the first number in the book and jump to that page. This way you make sure you are starting at a random point and not just at the beginning.
posted by Dr Dracator at 7:54 AM on February 4, 2016 [8 favorites]


To be honest ℤ is the most boring one-dimensional subspace of ℝ anyway.
posted by Elementary Penguin at 7:57 AM on February 4, 2016 [3 favorites]




#57361 Will Blow Your Mind!

Like usual with these lousy clickbait tactics, this turns out to be totally underwhelming: it's just an undistinguished instruction partway through the routine for computing EXP.
posted by Wolfdog at 8:25 AM on February 4, 2016


Oh, I get it now. 'Artisan' is the word you use when a middle-class person makes something a working-class person used to make... - @adamliaw on Twitter
posted by memebake at 8:27 AM on February 4, 2016 [8 favorites]


Why is it that all the good numbers are either prime or imaginary?
posted by petebest at 8:28 AM on February 4, 2016


What's Special About This Number?
posted by Nelson at 8:30 AM on February 4, 2016 [1 favorite]


Anyway, I can't tell you what the precise motivation for any given integer-as-a-service site is, but I can think of at least one application that needs good entropy, uniqueness, and integers rather than UUIDs. Namely, cryptographic salts.

I was into cryptographic salted caramel before it was cool.
posted by Rock Steady at 8:39 AM on February 4, 2016


To be honest ℤ is the most boring one-dimensional subspace of ℝ anyway.

Oh shit, we gonna get some number theorist beef in here?
posted by kmz at 8:51 AM on February 4, 2016


@kmz: I don't think so. Penguin is pretty much completely right.
posted by yeolcoatl at 9:10 AM on February 4, 2016 [1 favorite]


There's nothing wrong with mass-market integers, they're of uniform quality and consistency, and a good, dependable number. I prefer 3, myself, even if it is just another Anaxagoras-Bürgi brand.
posted by Slap*Happy at 9:37 AM on February 4, 2016


Now we know how many hipsters it takes to screw in a light bulb.
posted by rocket88 at 9:41 AM on February 4, 2016 [1 favorite]


I wanted 8675309, but that's already Jenny's #.
posted by OHenryPacey at 10:18 AM on February 4, 2016


Your favourite number succs.
posted by saturday_morning at 10:37 AM on February 4, 2016 [2 favorites]


so, basically they are trying to use something diffirent instead of UUID scheme because the UUID scheme provides strings of alphanumerical characters and dashes which are (supposedly)

UUIDs are integers, the alphanumerical stuff is just an encoding scheme (hexidecimal with some extra separators. It's sort of like with everyday numbers: 1,123,456 and 123456 and 1E240 are the same number, just encoded differently.
posted by cosmic.osmo at 11:44 AM on February 4, 2016 [1 favorite]


1,123,456 is an elegant representation, particularly in an appropriate font. 1E240 is awkward, mass produced. Functional, sure, but not artisanal.
posted by Nelson at 12:07 PM on February 4, 2016 [1 favorite]


Once they get up to the 10^9 range, they 'll have to admit to "Billions served", and you can't really get less hipster than that.

I wonder if there are any mathematically gifted synesthetics who can detect subtle differences in the sensation of large integers. I know the "5 is greenish-yellow and 10 is reddy-mauve" stuff, but can you get to "1033401693? A hint of rain on the nose and straw towards the middle of the finish, but it lapses into Plasticine on the out. Now, 39238849 is far better balanced...".
posted by Devonian at 1:03 PM on February 4, 2016 [1 favorite]


80085
posted by j_curiouser at 1:17 PM on February 4, 2016 [1 favorite]


Phhht.... artisanal? Handmade? Integers...

Damn hipster kids these days...

I prefer to get my numbers via radioactive decay and have them be truly unique...
posted by jkaczor at 1:26 PM on February 4, 2016


So, UUIDs are statistically unique. You call a function that generates a random value, and the result is highly, highly unlikely to match any other UUID generated by any other call to any other function that also generates UUIDs. But, there's no way to actually guarantee uniqueness unless you actually have a centralized service that says, "No, don't use that one, it's taken."

Which is more likely: your 128 bit random numbers collide with someone else's because you were unlucky while generating them, or because a guaranteed-unique approach suffered a rare checksum-indetectable hardware error during generation, transmission, & storage?

(I don't know the answer to that question)
posted by aubilenon at 2:55 PM on February 4, 2016 [1 favorite]


Sequential integers are more efficient in many databases but are only unique within the DB. The main reason some SQL people prefer unique incrementing integers is because they're a great way to build a clustered index where the table is physically and logically laid out from oldest record to last, because you end up using that integer for all sorts of table joining and queries (like a medical record or person number). Microsoft SQL server actually provides a mechanism for generating UUIDs in a sequential fashion. Since they are random and don't tell you anything about how one record relates to another in a logical (i.e. Chronological) way, if you make them your primary key and clustered index, you will literally be "shuffling" the table randomly for no good reason any time the index is rebuilt because the UUIDs are random and non sequential by default.
posted by aydeejones at 3:05 PM on February 4, 2016 [2 favorites]


I just want to apologize for my earlier comment. The integers are a one-dimensional submodule of the reals, not a subspace. Sorry.
posted by Elementary Penguin at 3:35 PM on February 4, 2016 [1 favorite]


Can this thing generate non-interesting integers? Not interesting ones, like 0 or 1 or 2.
posted by thelonius at 9:12 PM on February 4, 2016


I want mine to be perfect.
posted by nat at 9:40 PM on February 4, 2016


nat: How odd.
posted by vernondalhart at 12:38 AM on February 5, 2016 [1 favorite]


It's not just a joke, it's a matter of national importance for the United States. It powers critical infrastructure for the Smithsonian's design museum, Cooper Hewitt.
posted by e1presidente at 1:38 PM on February 5, 2016 [1 favorite]


« Older Going to the Puppies   |   Pentagon's Social Freezing Program Newer »


This thread has been archived and is closed to new comments