How Roman numerals broke the official dog database. 

"One of the main causes of the fall of the Roman Empire was that, lacking zero, they had no way to indicate successful termination of their C programs." -- Robert Firth

I, for 1, like Roman numerals.

Fun fact: You can name more than 37 dogs “;drop table *” without the database being full. (no legal advice)

As soon as the numbers came up in the allowed character list, I started imagining dogs named like vanity plates. You just have to name your 38th dog Skyl4b and you're good to go! After a number of generations with similarly constructed names, we'll eventually get around to naming them 5ky148.

"Assuming all atoms are the same breed of dog" - I wish more of my physics exams started with that assumption.

"The American Kennel Club used to keep track of dogs with the same name by using Roman numerals. They still do, but they used to, too." -Matt Hedberg

Usually, the AKC registered name isn't a name. It's a manufacturer serial number. It usually includes the name of the kennel the dog is from, a code indicating the litter it is from, a name that often has to fit some theme set by the kennel, and sometimes other weirdness. A quick Google gives me a picture of a dog named "Ableaim Patent Pending MC"

My database teacher would be mad for 5 different reasons if I handled things like they did.

This is why it's important to pick the correct data type when modelling a database

"They're registered by breed, and therefore antimatter is a different list" is not a conclusion I would've expected going into this video.

"if you allow multiple Ms, which some people do" - some people would include the Romans. Although III and IV were the most common representations for 3 and 4, you can also find IIV AND IIII.

The reality is, no dog registered with the kennel club is named “Spot”. They’ve all got names like Harry Potter secondary characters, although “Excelsior Ludlum Spottiswoode IV” might get called “Spot” at home

There's a slight issue with naming every atom in the universe with a 50 character name: there aren't enough atoms to register those names with.

I love the leap from “you could name every atom in the observable universe” to “everything can be dogs!”

Your solutions are nice, but I'm *way* more curious about why AKC's naming system has these limitations in the first place now. I'm trying to imagine a system where the roman numerals couldn't just be outright replaced with numbers, even pre-computerization.

My solution: IF one adds a space after a name, that doubles the number of names you can have. Therefore our number of names increases for every character not used under the limit. Such a system would probably annoy them greatly, however, so they would likely moderize the system.

As a database professional, this pains me a great deal. There is no reason (other than a lack of will) that they could not extend that character limit, and absolutely no reason why they should not.

Writing 3999 as MMMIM instead of MMMCMXCIX, as it should, made me shiver. But then you went ahead and proposed writing 3 as IIV. That's the most cursed thing I've ever seen in my life.

Given how dogs are named in the AKC, it's unlikely to see 38 of the same name anyway. The majority of dogs official AKC names include the kennel names where the dog was bred, and potentially the kennel name of the borrowed sire.

One thing I find interesting about Option 2 is that the word for 18 in Latin is "duodeviginti", literally meaning two from twenty. So by all rights it should be written XIIX.

Going to guess the database was using Hollerith cards, like those that used to be used for census data. 80 columns of 10 digits, so likely they allocated 2 rows to store a representation of the Roman numerals, using 3 bits, and then 3 sets of numbers, and a check bit, over the 2 rows. Leaves 78 columns for use, with likely one taken for breed (512 breeds should be enough to handle any likely future use), leaving 77 for name of dog and a reference field to another card, containing address. Used machines that at the time were common, easy to get hold of as used and working, and also had the whole "computerised" feel to it. Easy to do lookup per field as to what would be allowed, and what translation to use, when sending to output as well.

I love how excited Matt is about Option III but never even mentions how it only allows II more dogs.

Veritasium just posted a video about number 37... this is so unbelievable... how I supposed to live with that ???

"Assuming all atoms are technically the same breed. Antimatter. That's a different list" that one got me good

The rule about "no more than three in a row" and, indeed, the use of negatives is a partially modern invention. The Romans did regularly have to deal with numbers over 3999-the classic legion was larger than that, never mind the census data-and even older modern clocks have digits like VIIII instead of IX. There's no reason we can't update the rules further.

I thought of option 2 while watching the video, a couple minutes before you showed it, that felt awesome lol

I absolutely love how Matt never once touches on the obvious solution, yet everyone knows it's there. That's the comedy here.

Why has Skylab been absent from your previous videos? The road to two million is paved with dog treats.

There's also the 4th and possibly most boring/"intellectually challenging" option depending on who you might ask: migrate the database to a more modern standard c;

I was wondering where I'd seen this video before, up until you plugged the podcast. Say hi to Bec!

I'm pretty sure that the ultimate reason for this limitation is that the original system was likely coded in COBOL. COBOL has this thing for building data records using ASCII characters with a fixed width for each field, with a newline after each record. This makes reading through a file with a buttload of records in it quite easy since every record is exactly the same number of bytes, and finding a particular value within a record is easy for COBOL because it's always in exactly the same spot. This is what made/makes COBOL so good for this kind of record storage. Today, we'd use some kind of RDBMS but back when computers were new(-ish), such things didn't exist yet, and flat files written to tapes were the thing to use, so COBOL and the COBOL way of storing data made a ton of sense. In particular, knowing the size of each record in advance let you scroll through a tape very quickly and easily. This is important with tape since rewinding a tape is very slow, so overshooting the mark was very bad, and you couldn't just read the beginning and end of a file at the same time with a tape (which is needed for many modern storage methods, like ZIP files and RDBMSs), you had to read the whole thing from beginning to end. You also had to write the files from beginning to end without going backward, and it was unlikely that you'd be able to store the whole thing in memory at once, so you couldn't just make a table of contents for the beginning and then write out the rest of the file. You couldn't have generated the table of contents until after the whole thing was written so it would have to be written at the end, but you need the table of contents to navigate, so it needs to be at the beginning, but you can't do that, and OMG, it's not going to work. That's why you didn't usually put ZIP files onto tapes in the early days of ZIP files; their table of contents is at the end of the file. Tar files put it at the beginning, but at the cost of having to have the files already compiled onto a disk of some sort as an intermediate. The individual files could contain whatever tables of contents were necessary for navigation and because you were using a disk, you could write them to the beginning of the file after you'd already written the rest of the file's contents to disk. But you can't do that when you're writing directly to a tape with no disk storage intermediate. COBOL and its fixed width record-based files was perfect for the era when you had tapes but no disks.

How we began: In America, only 37 dogs of the same breed can have the same name. How we ended: We could potentially register every observable atom in the universe in the Kennel Club, even if they were the same breed of atom.

Matt: "And that's where most videos would end: here's a crazy fact, here's the reason behind it, job done. But not here! The Stand-up Maths policy is to try and fix things! And I reckon I've got some solutions so Americans can have more dogs with the same name." Me: "Arabic numerals. With six digits, you can count op to 999 999." Option 3: if you were to start at 39 instead of at 1, you can have two more dogs with the same name. I like this channel so much.

When Skylab taps Matt’s arm to ask for another treat ❤️

Or, Option 4: The ACC drops Roman numerals INSIDE the database, and uses either base 10 (999,999 Dogs of the same breed with the same name) or base 16 (0xFF FFFF (16,777,215 dogs of the same breed with the same name). To keep their Display in Roman(ish) numerals, they could have a 'look-up table' in their database to pull up a string based on the value, adding new Roman(ish) numerals. We already have I=1 V=5 X=10 L=50 C=100 D=500 M=1,000, so, borrowing from Engineering notation, we can add T=5,000, P=10,000 E=50,000 Z=100,000 Y=50,000 ... Here the Engineering prefixes stop, so the 'No More than 3' rule and the 'no duplicate 5s' rule leaves us with ZZZY, 350,000. So we go with another Engineering habit, "when the 'numbers' run out, start using letters." So we'll have A=1,000,000 B=5,000,000 F=10,000,000, and now we can represent all 16x10^6 dogs with Roman(ish) numerals. Although it might just be easier to just use the hexadecimal numbers, the database is already working in that base internally (actually, it works in base 2, but 4 hex digits completely describes a 16-bit 'word,' the standard memory size in modern computers.)

First thought was "what if they allowed an additional digit? How far can you get before hitting 8 numerals?" I was quite pleased to realise that 88 is actually the first Roman numeral of length 8 😁

Option 4: The American Kennel Club stops acting like an elitist clique and just uses a sensible database based on unique ID numbers....

This one of my favs from you now.

Would you believe this is the first thing I thought of after Veritasium’s new video about the number 37?

This might be less of a common issue than it might seem at first. AKC breeding dogs have 2 names, their "common" name ("Skylab"), but then they have these crazy gelical names like "Parker's Dazzling Stellar Cosmodog". So the AKC registered name tends to be very unique anyways.

Matt writing 3 as IIV is the most characteristically unnecessary change and im a fan of it

Hearing your podcast now I want to come back to the Pizza 🍕 problem of the second (?) Episode. You tested pizza and said that the center of the slices was off. And I wonder: how much of can be tolerated? Is there a distribution mechanism between 2 friends that compensates the off center part? How does the ratio between crust length an area change? So many math questions with off-center sliced pizza I would love to see answered here on this channel!

Brilliant :) Made my day !

Fun fact: there is actually precedent for usage of option II in Roman times; the Wikipedia article on Roman numerals specifically mentions IIX for 8 and variants as "Irregular subtractive notation" (not necessarily consistantly - it shows both IIXX and XIIX for 18, for example).

Even binary, the least dense numeric base, has more room with six digits than roman numerals It also has zero, so you can make use of all 64 rather than just 63

what a nice dog) love that moment when Skylab asked for one more piece of food)) "hey, human, you forgot one more piece to feed me"

Yoo tnx dude, everything works. I LIKE IT

Another option is to start the list at -27 (-XXVII) and go up to 37, which gives a run of 65 different choices

I can't believe I was right in the middle of a deep dive on Roman history, Roman mathematics and the Roman calendar, and look who drops into my feed! And Skylab too! 🐕

Thanks for your videos

After some Googling, it seems like the Romans actually had signs for the 1000s, like ↁ (5000) and ↂ (10000). But they had multiple ways of indicating 1000s. An alternative notation, and one that is easier to grasp, is the vinculum notation, where V̅ would be 5000. I'm not sure whether you could actually say that Roman numerals don't have the option to go above 3999. They could, but because there were multiple methods of handling that, we don't seem to teach them nowadays.

Gotta love that he didn't once mention the solution of using regular digits instead of roman numerals because that would be too easy

Regarding 8 as IIX the 18th legion used XIIX as their label which precisely follow that rule you mention. Yes, it was unusual but XIIX is nicely symmetric and so was probably the reason why it was chosen. Also of note is that it is not only read the same forward and backward but also upside down and that makes it very nice for a label which may appear in many directions during a battle depending on the angle of the banner with the ground.

Oh goodie, now I want all of my atoms to be named individually and registered. And every morning there has to be a roll call and we'll welcome new participants and hold a short memorial for the departed...

This video just kept. Getting. Better.

Fun fact, I used to had a dog breeder as a client, and would see the bills for the name registrations. Most of the time the dog's registered name would be [Breeder name] [Actual Name]. Now, this was in Canada, so it was the Canadian Kennel Club, so their naming rules may be different. So if the breeder's business was called 'Starwood Country Kennel' they would register their dogs as Starwood Rover, or Starwood Spot. That way you never have to actually worry about having dogs with the same name.

I was here for the invention of Parker Numerals

What the hell was the conclusion?!! I LOVE IT! 🤣🤣🤣

8:45 I once was at some friend's house, when I noticed that the clock they had hanging in the kitchen had all of its values written in Roman. Nothing out of the ordinary, classic stylistic move. But, I noticed that the number 8 was written the Parker Way, in other words IIX. Ever since then, I was confused as to what was the real deal about Roman numerals

"No one can argue with that!.... people will argue with that :(" It's really nice to see how self aware you are :'D great video

"They have so overpowered the rest of the name that there are enough names for everything!" Oh, Matt, how could you apply combinatorics to the characters in the names but not to the atoms you're trying to name? Ok, so we name every atom, but we can also name, e.g., "that galaxy", and presumably things like people and, well, dogs. But what if I want a name for me and my dog? What if I want a name for everything in my galaxy except my dog? At the end of the day, we need names for every combination of atoms, and that's going to need rather a lot of names. Loved the video.

Here are a couple of other solutions: 1. Just expand the database field to use more characters. This is a simple thing to do (depending on db system) 2. Use an integer instead of a character based Roman numeral. 3. Ok so you want to keep the Roman numbers, covert the database to use an integer and on the display of that value convert it on the fly to a Roman numeral. This could be done on a db "View" (assuming a sql type db) or on the UI.

Matt: I’ve innovated Roman numerals with IIX! Meanwhile 9 y/o me thinking of it in Latin

Note on option 1: there was no actual rule "only three of each in a row" to the people for whom "roman numerals" were just "numerals". Even back when Julius Caesar still had things to say, IIII instead of IV could be found in official records.

The delivery of "... antimatter... that's, that's a completely different list" was gold

As he's going through all of these proposals I'm just thinking to myself: Convert to Arabic numerals. They would allow just under 1 million dogs to share the same name, wouldn't they? Better yet: Cistercian numerals. 6 digits of Cistercian numerals would allow 999,400,150,000,000,000,000,000 dogs to share the same name (That's nine hundred ninety-nine sextillion four hundred quintillion one hundred fifty quadrillion dogs with the same name, by the way.)

Option 5: The original database was probably restricted to alphabetic characters only (no numbers). They could have used 6 letters instead of Roman numbers for millions of combinations.

Or you know, they could store the numbers as a database defined integer, and have a function convert said integer into a roman numeral for display. The limit at that point is the size of an integer in your DB, which should be easily large enough.

We had a regex at work that only allowed up to four characters. It went for years before someone realized that it was missing 18 and a bunch of other ones. I think the fact that the gaps weren't consecutive made it easier to miss.

damn, i love option 3. this was amazing.

Love it... thank you!!!

Some dogs are made of multiple atoms [citation needed] so we really need names long enough to handle the power set of all atoms in the observable universe.

11:36 "I have become very focused on the Roman Numeral aspect -- don't get me wrong, had a great time." Having a great time with mathematics and sharing it with the world is what makes this channel awesome. :)

You can name the all the individual atoms, but if you start coming up with names for galaxies, etc. you're naming every possible set of atoms. And that's a MUCH larger number.

I was on board with Option 2, but now I’m all in on Option 3!

My partner related this story to me [web developer] while we were out walking our own dog last night. Never has a database design story caused me more psychic damage.

"NO ONE CAN ARGUE WITH THIS. people are gonna argue with this." i love it XD

IIV for 3 is so incredibly cursed, thank you so much

"Anti-matter! That's a different..." (pause for thinking)

The "changing the sign" explanation makes perfect sense to me, and parsing is quite a bit easier with that in mind (you would only need an accumulator, a mapper, and the previous individual value). But Option 2 would make that parser a bit more difficult to implement.

missed opportunity to name the dog MattLab...

I have to admit that I've never heard a more confusing explanation of Roman numerals in my life. I've been calculating in them for over 50yrs and now I'm confused 🤣

"No one can argue with that! People are going to argue with that." I need this as a reaction meme

Or convert all the current Roman numerals to base 10 strings in the database (since it currently only accepts strings not integers) and display them as Roman numerals when retrieved. For example convert “XXXVII” to “37” in the database. That gives you 999,999 of each name.

So THAT'S why dogs are registered with such weird names. Growing up, we had a Great Dane named Rufus. Even though he was not registered with the AKC (at least I don't think he was), my mom insisted that his "official" name was Sir Rufusson of... "(something, something, something - I don't remember because we just called him Rufus).

That was GENIUS!!

Option two is by far the best

I have been advocating for Parker Roman Numerals for nearly 3 decades now and I am glad that I'm no longer alone in this fight. Welcome, fellows.

i love how mathematicians go on these weirdly absurd thought experiments, meanwhile the AKC probably hasn't even used roman numerals in decades and the rule only exists because it's been grandfathered in.

That was fun. Being a geezer learned about Roman Numerals in grade school. I can't imagine using them for construction or engineering.

Matt!! You told us the "rule" was: if a character has a bigger numeral to its right, it's negative... IIX does not work! The first "I" does not have a bigger numeral directly right off it

I like the idea that "elementary particle" is technically a breed of dog, and the entire observable universe is technically America, so each particle should have its own unique name registered with the American Kennel Club.

In No Man Sky, I would always name each type of planet with a roman numeral system up to 30, then switch to decimal after 30. It was a little confusing with each numeral sometime in the teens so that why I decided on the switch.

By now I've given up on understanding how Matt thinks, and I'm completely fine with it. But the last thing I was expecting when I clicked on this video was that it would end with anything related to antimatter.

Pick a random number between 1 and 100.

Another question: Is this really an issue? Here in Germany people who register their dogs use fancy unique names like "Harold Vincent von der Vogelweyde" (the "last name" usually referring to the dog's sire) but call them "Spot" at home.

Skylab is such a beautiful pup. Labs are the best!

I am LIKING, SUBSCRIBING, and ringing the BELL, because you did not ask me right up front. By the way, very clever and funny. Thank you.

Skylab. :) Golden retriever here. Considered naming him Au thinking it would be great fun to call out 'Eh! You!" at the park. Decided to keep the au in the name and went with Tau. :)

Can't have too many "M"s or you'd end up with that Crash Test Dummies song.