Hello, and welcome to another episode of The Bike Shed, a weekly podcast from your friends at ThoughtBot about developing great software. I'm Joelle Kinville, and today I'm joined with fellow thoughtbotter Sally Hall. Hello. And together we're here to share a bit of what we've learned along the way. So Sally, what's new in your world? This feels like a silly thing to be new for me, but I read a lot, and mostly in the form of audiobooks, which I will not hear any arguments, but that's not reading, it is. It is absolutely reading. Yeah. Recently I've just started reading actual physical paper books again. I sort of stumbled into a bookstore, needed some time to kill, bought a couple books, and it's been nice to sort of go back to the original way that I was reading, and it's...
Definitely a more focused way to read than audiobooks where I'm often listening and knitting or listening and cleaning to just sit and focus completely on what I'm reading and the smell of the book. All the things that I've always loved about paper books, I'm sort of rediscovering. That's always a tough trade-off, right? You get more volume if you're able to multitask. but maybe you get less focus. Yeah, and you tend to, I don't know, I tend to get a little distracted and sometimes realize like, oh, I just missed something. I guess I need to rewind and something important happened, which, I mean, I certainly can get distracted while reading a physical book too. It just happens less. It's also nice to know how everything is spelled. It kind of drives me crazy. I read a lot of historical fiction and there will be names that I've never heard before and I have no idea how it's spelled. I'm sure. Any favorite books or series that you would recommend in that genre? I really like Ken Follett. He wrote Pillars of the Earth, but has written a lot of other historical fiction. He does like...
epic trilogies. The audiobooks are like 40 hours long, but they're really good and really interesting. I never liked history when I was in school, but now reading historical fiction, I'm like, oh, there's more to history than just politics and war. Like learning how people lived and things like that is really fascinating. It's more than just names and dates. Yeah. Yeah. It's human stories. I feel like sometimes historical fiction and fantasy authors do this as well. They just have massive casts of characters and you end up having to, you don't have to memorize dates as much, but you do have to memorize an awful lot of names. Yeah. How has that been in some of his work?
Oh, that's interesting because definitely there's one book that, and I get the names of the books mixed up, but I can remember which century they're in. So whichever one he wrote that is in like... right around the year 1000. I think it might be called the evening in the morning. There are families where like, I guess at that time, there was a custom of giving your kids names that all start with the same first letter, which I find even more frustrating because it's like, okay, these three brothers and like, one of them is kind of a good guy and the other two are kind of like bumbling idiots, but their names all start with the same letter and I cannot remember which is which until they start talking. That's definitely not a trend that we still do today. Oh, absolutely.
My kids are twins, and I worked really hard to make sure I didn't do that. I figured they're already going to be sharing so much, and they might as well get to have their own initials. Right, right. You want them to have their own identities. Yeah. And then it's easier to write on a basketball, this belongs to G and this belongs to H. That definitely makes things easier. Yeah. Well, speaking of things that are still current today, but that actually are surprisingly also done in the past, you and I were in an interesting conversation recently where we were talking about...
The idea of a lot of what modern apps do for us and how people used to do them in the past. And a lot of these things people were able to do with what I can almost describe as paper data structures. Yeah. So instead of things like using a linked list or an array or something like that, you might use a card catalog or you might use a physical index. A lot of these things we think are unique and new to the digital world actually had pretty clever solutions to sort of information management in the analog world.
Yeah. And I mentioned that this is, you know, if I wanted to really, really simplify it, it's basically what my master's degree is. I went to school for information science. It was at the School of Information and Library Science. So it was like a combat school, but I focused on the information science. And it really was just a lot of like, how do we organize and access information broadly? So not necessarily within the context of a library or within the context of a computer, but just like, you know, our world is full of information and we need to access it and we need to store things. And how do we do that all the different ways, which was really fun to think about.
I love that this is the thing that people study at a more abstract level and not just in the context of computers or digital solutions. And I remember one project my group and I worked on, like, how would we organize a collection of board games? Like, what are the different ways that someone would want to retrieve a board game? You know, like, there's the basic ideas that, like, okay, you want to be able to look at it by title, right? What gives me Settlers of Catan. But when you get really deep into it, there are people who...
are interested in board games created by a specific person or from a certain country. There's, you know, German board games are a very specific type that's not necessarily the same as, say, Japanese board games or the art or the theme or the mechanics and all of that. And just thinking about, like, the information that we're looking at is a board game, which I guess, you know, information might be kind of a stretch for that word, but I think of it's information. It's a set of rules about how to play something. It's a list of board games and you could absolutely store that in a digital system, in a database. Yeah. With a bunch of columns and maybe you could sort or search by these different columns.
How do you do that in a situation where you're not using a computer? Yeah. So I think my absolute favorite organizational system is a card catalog. I actually have a tattoo that is like the handle of a card catalog drawer. Oh, I love that. Yeah. And I think what I love about a card catalog or the idea of it is that it's not a one-to-one where there's like one piece of paper represents one book or board game or something, right? Often you'll have either mixed together or in separate cabinets, you'll be able to look something up by author or by title or by subject or by something else. And the idea is that you go through and you look for like, I don't know, dogs.
And then you'll find all the cards that represent a book or a board game about dogs. But you can also, in another cabinet, look up by title, Old Yeller. And so the book for Old Yeller is going to have a card in the dog section under D, but it's also going to have a card in the section for titles that start with O. And so I think a lot of times when we think about... using paper to organize or track things, it feels like it has to be very one-to-one, like where's the card for this book? Right. But really you can get more expansive in that and you can, you know, address it from different directions. And there's not just the one way you can find this book. There are lots of angles that you can access it. Every dimension can get its own, I guess, box of cards. Yeah. Yeah. And so, and you can have them separate or you can have them together sort of depending on how you find it. And one of the things that I really love about physical systems like that is that they allow for browsing and sort of
being surprised by what you find in a way that is a little bit harder to do than digitally, right? If I'm looking for a board game digitally, I know that there are several websites where I can go and put in all the attributes I want for the board game and it will show me the one I'm looking for or a couple that are sort of like the one I'm looking for. But I really enjoy opportunities to sort of happen upon things, right? So if I'm looking for a board game and I've got like a little card catalog and I know I want it to be about trains, I'll go find the train section of subjects. But then I might also, while I'm physically flipping through those cards to find the one I'm looking for, happen upon a couple of other ones that a computer would have filtered out for me.
but I get to stumble onto them and discover new information in ways like that. Yeah, it's like the contrast between the idea of browsing versus the idea of searching. Yeah, and when something is in a physical data structure, it's a little easier to do both at the same time. You know, I'm searching for one specific thing, but I can sort of peripherally browse at the same time. And I find that digital systems often, you kind of have to choose one or the other. If I'm searching, it's going to be pinpointed. I've had a lot of conversations with people about like, how do we replicate the experience of going into a library, looking for a book, you find the book you want, and then because you're standing in front of the shelf, you also see other books you wouldn't have thought to look for that are nearby.
Which, I mean, you can do digitally. I just feel like there's more joy and surprise physically for me, you know, on paper. That makes a lot of sense. I'm curious, you're sort of talking about having a cabinet or a drawer for each attribute. If I'm designing a database table for board games, I might think about normalization. And maybe authors are stored in a separate table. Maybe country of origin is also a separate table. Things like that. Are those sorts of concerns something that you would do if you were designing a...
card catalog library? Like is normalization something that goes back into the analog age? Oh, that's really interesting. I mean, I think to some degree, it could. Like if we're talking about, you know, country of origin, we might have, you know, the drawers, the drawer that's sorted by country of origin. And then within that, little dividers for each country. So not necessarily writing.
Japan at the top of every card that is for a record from Japan, but having a divider that says Japan and then knowing everything from that to the next index card or next divider is Japan. So you don't have to look at each one individually. I feel like that would be a little bit like normalization. But I think there's definitely still the idea of like, if we have, you know, three different drawers for different attributes, ways you can look things up, you're still going to have all the relevant information for actually retrieving that book on all of those cards, right? So we're definitely going to be repeating some stuff. And you might have information about the author on the card, even though we could theoretically have a separate author drawer where this has the author number on it. Then you can go look up details about the author in that drawer.
So I guess my answer is I don't know. Sort of? It sounds like there's a lot of the same concerns you would have with a digital database where full normalization sometimes is inconvenient. And so sometimes we don't go all the way. Yeah, yeah, exactly. But also I can imagine some of the techniques you've described would make it easier if some of the data changes and you have a single source of truth. Yeah, and in similar ways, it's, you know, the less normalized it is, the more space it takes up and the more work you have to do to update or change things. But then perhaps it makes it easier to retrieve information. So you got to sort of...
You don't have to jump between drawers as much. Right. When you're looking for it. But when you're updating it, you do. Well, if it's not split across multiple drawers, then yes, you have to update it in three or four places. Yeah. It's possible that if you split it across multiple drawers, you would only have a single place to update because you've been very detailed about this is the author's drawers. And so if I have to change the name, I change it in a single place on this one card. Yeah. But now when I want to retrieve it, I have to pull a card from the country drawer and the author drawer and the game drawer. And that's getting really annoying. Yeah. That's really interesting. I hadn't thought about normalizing a...
a paper data structure before. I imagine pulling from different drawers is almost like doing a join in a SQL query. Yeah. Unlike a SQL query, though, when you're done, you've got to put those cards back in the right drawers. That's true. That's true. So, you know, there are uses for computers still. I almost wonder, too, the way you're describing different drawers that are sort of based around attributes, it feels a little bit more like how I've heard people describe graph databases, where things are more about describing attributes and relationships and things sort of relate to each other via IDs, and you're building out a graph more than like a table structure. That's interesting. I don't know very much about graph.
databases. Can you tell me more about them? You're hitting close to the limit of knowledge that I have of them. Yeah. I mean, I think about, so I studied math in college and say the idea of like a mathematical, like the idea of a graph where there are vertices and edges and things connect in different ways, I think definitely corresponds to sort of the way we're talking about this information storage that like different shared attributes and sort of the way you can follow a trail of information. I was actually watching Celebrity Jeopardy the other day and Roy Wood Jr., who is a...
amazing comedian, was on there and was talking about having an encyclopedia in his room when he was a kid. And that when he got in trouble and sent to his room, the only thing he could do, because we didn't have phones and stuff back then, or as much digital entertainment, but he had a set of encyclopedias. So he would just sit and read through the encyclopedias. And he said that the encyclopedias he had would have like, you know, he'd read a section on, read the entry on the Jaguar. And then at the end of it, it'd be like, you might also want to read about tigers. And so then he'd go read the part about tigers. And so it's interesting to me because it felt similar to the idea of going down like a Wikipedia rabbit hole where you look something up and then you click a link and it takes you to another thing. And the idea that like people were doing that.
in books before we had wikipedias and links like you know it's harder to go you know from the j volume to the t volume to get from jaguars to tigers and looking up tigers by having to remember how it's spelled and all of that instead of just clicking a link but it's really the same idea of like sort of just chasing a trail of information through links and references so encyclopedias are interesting in that they're kind of fixed they're not like a coward catalog where you can reorder things or add more information in them right they're a fixed sort of printed list of things and they're ordered sequentially. And so if you're trying to find something, you have to, I mean, they're sorted, so you can do effectively binary search on them to find something alphabetically.
In my mind, I'm almost getting parallels with the idea of an array as a data structure, sort of contiguous blocks of memory. Yeah. Is that a fair comparison, would you say? Card catalogs are maybe a little bit more like a linked list or even a database, whereas something like an encyclopedia or maybe even a phone book is more like an array. Yeah. And maybe like an array of hashes, but yeah. Right. Because there are multiple chunks of data within a position. Yeah. And there are multiple attributes for each chunk that's in the array. Right. But yeah, I think that's really interesting that it's like, it is ordered. There's not really a great way to get from one item to the other beyond just...
going through them or randomly choosing one and going from there. I guess with a physical book, you get the opportunity to just sort of like flip through pages and land on things. But then there's, you know, the idea of an index in the back of a book. Right. And I think cookbook indexes are some of the most interesting ones because you can look up a recipe by the name of the recipe or the category or ingredients in it if you've got a good cookbook. And so then what is that like? Is that like a, I guess it's like a database index. Right. We just straight up lifted that from the physical world and brought it into the digital one. We just stole that name. Yeah. Like, oh, you might want to.
You're going to store things physically in a certain order because that's most efficient for storage. But for lookups, we might need to have the data sorted by name or by ingredient or whatever. What if we stored that as a separate chunk of data somewhere else, just the sorting? And that made our lookups faster. Yeah.
And indexes were books and pre-software were all made by hand. So someone would go through and for each recipe, go like, okay, what do we need to index this as? We need to put this in the index under the name of the recipe. We need to put it under the ingredients, et cetera. And this is actually making me realize that this is sort of something that I did my master's thesis on, was the idea of comparing a human-made index for a set of data to a machine-made index for the same set of data. So I took one of the data sets that's popular to use for information science research because it is huge and diverse and freely available is...
all the documents from the tobacco lawsuits in the 90s. Okay. Yeah, I can't remember exactly what it's called, but that is a huge data set that you can access and do stuff with. And so it's all different kinds of documents, you know, like memos, emails, research papers, stuff like that. And that has an index that was created by humans. So I took some queries and did searches with the human index, and then I...
created an index using software based on like keywords and stuff. And then compared the two for, is the computer generated index as effective as the human one? Because it certainly is cheaper to do if I could do it on my own as part of a master's thing, rather than, you know, I'm not sure how many people they had involved in indexing it by hand originally. And I think what I found was that the human index was better. And so certainly, you know, if I was doing...
more in-depth research, I could see, like, okay, well, that was one automated index. How does it compare to other kinds? How can we improve to get one that was as good as the human one? It would be interesting, and it's probably hard to know because the data set is so huge, sort of what are the failings of the human index? Are there some things that the human index misses? Yeah, and it would be interesting to also approach it, you know, like, decades later, because it has been decades since I was in grad school.
to sort of see like, are there differences in the types of searches people might do? You know, maybe for the types of queries that I was working with 15 years ago, the human index was better. But now that we've spent a couple more decades asking computers questions rather than... only people, have we learned to ask questions in a different way that might make the automated index better? I feel like that right there is probably a social science research paper. Yeah. How have humans changed the way we ask questions in the decades since computer searches have become broadly available? Yeah. And does it change how we want to organize information? And then this isn't even considering AI, right? Right. Like, could an LLM build a good index for a giant probe of data? Yeah. I know. I think the thing that I get stuck on the most with AI is that it sounds very confident, but like, is it right? Yeah.
But that's something a paper could probably compare if you have a human-made index as a reference point. Yeah. And that is one of the measures for the indexes that I compared was the idea of accuracy and precision, which are like different things, right? So accuracy is like, is this correct or not? And precision is like how granularly close to what I'm looking for is it? Right. So yeah, it'd be really interesting to compare. Yeah, I'm thinking about it. And I feel like my experiences with something like ChatGPT is that they're not terribly accurate, but they can be pretty precise if you ask them questions in the right way. So I'm curious now sort of what makes for a good index and how does a human go about that?
I can imagine you probably have to start by thinking about a couple of dimensions that you care about indexing by. So maybe you're indexing by keywords, but you'd mentioned that for something like a recipe, you might want to index by ingredient. I imagine for something like lawsuits, you might want to index by participants or plaintiffs or something like that. Maybe you want to index by location. Yeah. And you may want standardized terms within all of those. Right. So like if I'm indexing by location, when I'm indexing this book and I put North Carolina and then you're indexing your book and you put NC. Right. Right. Those aren't going to line up together. And now this is sort of how we ended up with the Dewey Decimal System at the Library of Congress cataloging system is the idea that like, you know, if we break down everything we think something could be about.
and assign it numbers and it's easier to organize things by those subjects. But also within that, like not just, you know, because ultimately a book does have like one Library of Congress number. It's not going to be on four different shelves at the same time because it's one book. But that there are standardized topics and types of books in the Library of Congress system. So all the books that are historical fiction are going to be called, for example, and I am not an expert in the Library of Congress system at all, but like historical fiction, right? Instead of novels set in the past or something else that you could call it.
So there's a sort of standardized set of book categories that the Library of Congress manages. And it's interesting because there's inherent bias in those. Just by creating a set of here are the categories of books that we think are important. Right. What do you omit? What's more granular is going to be biased based on your worldview. Is that set fixed? Or does the Library of Congress sort of update the categories every decade or so? I'm not sure, but I would imagine that things get updated and things get added. Like the actual classification system as far as where books live on the shelf.
I think doesn't change very much. But I think as far as adding new categories, certainly there's got to be a category about software. But 100 years ago, there wasn't software. So there wasn't a category about software. I imagine things like sci-fi or fantasy or things like that didn't really exist much back then. Exactly. And there's historically been issues with the two main systems that we use in the US in libraries, the Dewey Decimal System and the Library of Congress, both have some pretty problematic biases in them. The range of numbers allotted to books about the Bible is huge, whereas the range of numbers allotted to books about...
literally every other religion in the world are sort of all clumped together. And it's like, we care about very specific aspects of like, this is this kind of book about the Bible and this is this kind of book about the Bible. Whereas like the Quran, the Torah, other holy books that I don't know about are all just like other religious texts. So it's very othering. It's very centered around what felt important to the people who were in charge at the time it was made.
That's interesting because it's not only the sort of importance you give something by giving it a dedicated category, but there's also a number space that goes with it. You get certain amount of numbers you can assign. Does that mean that the system can run out of numbers for a category? No, I don't think so, because I think it's a little bit like software versioning. You know, just stick a dot on there and add another number. Okay. A little bit to some degree. So it's not like IPv4 where you're like, oh no, this block of IP addresses is running out of numbers. No, no, because there's not like a standard length. And as a book becomes more specific, the number gets longer because it's like Q is science and Q2 is biology and Q21 is plant biology and Q213 is the biology of poison ivy. I don't know if this is right at all other than I'm pretty sure Q is science.
So you can just keep expanding the number to make it more and more specific. That makes sense. So at least they designed it to be expandable. Yeah. Yeah, it is expandable. It's more about what topics got granularity in the high-level namespaces versus which ones you have to go really deep. Right. Yeah. I feel like this is a little bit like the prestige you could get from having a single-letter username on Twitter back in the day. It shows that you were early to the game. Yeah. It's almost the idea that if you have a very short number for your book and your book is rather specific, that means that, I guess, back in the day, your field was a big deal.
Yeah, I think you could say it that way. And it's, I don't know, I think about bias and things like this a lot. And the idea that, you know, people sometimes think that because a computer is doing something, it is therefore unbiased because it is a computer. And how could a computer have bias? Whereas we know that algorithms inherently have the bias of the people who created them, right? If you're assigning value to something, you've got to code into the algorithm what value is good and what value is bad. So you get out what you put in. And if your input data is biased, then that's when you impact your output.
Exactly. And all of our input data is biased because we are humans and everything we do is biased. So similar to, you know, if you're working with chat GPT, you've got to analyze it with the awareness that like, this is not an objective system. And so let me take into account the context in which this question was asked in the context of the information it's got that it's used to create the answer as part of evaluating it, you know, instead of it just being like, well, it said golden retrievers are the best dog. So obviously they are. And that's something that we've been, I think that's interesting, but like, that's not a new problem at all. Right. That's baked into our organization of.
libraries as long as we've been organizing libraries. Is that a problem that is studied as part of information science? It wasn't a huge part of my time in information science that I remember, but I do know that it's always been talked about the idea of librarians are some of the fiercest activists for freedom of information and freedom of speech and things like that. So librarians and by association, information scientists do tend to think a lot about like, okay, what is the way that we are using and starting retrieving and allowing people access to this information doing to the world? So the social implications of the information structure. Yeah.
So we've talked about a variety of physical-ish data structures. We've talked about card catalogs. We've talked about sort of reference books. We've talked about indexes. Are there some other physical ways of organizing information that you really like? I really enjoy Rolodexes, which is probably just because they look cool. So for our listeners who have never encountered Rolodex, can you explain what these are? Yes.
For those of you who did not, you know, grow up in the 1900s, a Rolodex is generally a way of, I think it's most often used as contact information. So it'd be like an address book. But instead of it being like a fixed address book of pages, it is each contact is on a card and then they're all stuck together in this thing that you can sort of spin to look through them. So it's sort of like imagine a card catalog drawer, but instead of it being a flat thing, it is circular to make it easy to sort of like flip through them and to look at one without having to remove it. Right.
If you've seen Mad Men, you've probably seen someone use a Rolodex. And they are sorted by some attribute, probably alphabetically by name or something like that. Yeah. And I guess the advantage there is that it's not fixed so you can insert new records. And remove records without having to erase or, I mean, you could theoretically run out of room. But if you have room on the Rolodex as a whole, then you have room in any spot. You're not going to be like, well, I've got no room left in the R's. Where do I put Ronald? There's always going to be, you know, you can stick him in wherever. So in some ways, it's almost like the sort of phone book we described earlier, which is almost like an array.
Except that this one's much more expandable. It's easier to just add something in the middle. Yeah. It's almost like a single dimension card catalog. That you can insert things into anywhere instead of just at the end or at the beginning. And you can access them anywhere instead of having to start at the end or at the beginning. They also just look cool. That's an important attribute. Yeah. Is your organization system, is your data structure fun to use? Does that matter? We need something like big O notation, but like big cool notation. Yeah, that's CN squared. It's very fun.
So in terms of like its access properties, initially I was like, oh, this is almost like a hash map and that you can arbitrarily insert and remove from places. But that's not quite true because you do have to sort of browse through it to find things alphabetically. Yeah. It is more like a sorted array. Yeah. It's really ultimately kind of just like a cool container for a sorted array.
Yeah, yeah. But then there's this other kind of antique address book. I don't know if you've seen these, but it's like flat in a case. And there's a few pages for each letter of the alphabet. And when it's closed, there's a little slider that you can slide to a particular letter. And then when you open it, it will open to the beginning of that letter section. Oh, that's fun. So you don't need to sort of do the binary search thing where you're like, oh, let's see. Oh, a little too far in the ends. I need to back up. Oh, now I'm in the J's. Yeah. Like once you're there, you do have to do that. But it gives you that first step of jumping to the letter that you're looking for. I feel like tables are a thing that we've sort of always really valued.
It's like tables written out on paper. Before the age of the spreadsheet, we had something kind of like spreadsheets on paper. All sorts of accounting things or just balancing a checkbook or all of those things were done effectively as tables. Yeah. The very first catalog of the Library of Congress was a table where they listed the title, the author, where the book came from.
and the cost of the book, because if you didn't return it, you had to buy a new one. Oh, that's funny. Yeah. Can you still borrow books from the Library of Congress? Well, I don't know for sure. I didn't finish reading my Library of Congress book, but originally it was a library for Congress. So not anyone could borrow a book, but members of Congress could. I see. Okay. But you're on the hook if you don't return the book. If you borrow an expensive book and you don't bring it back, we know exactly how much we paid for it. And, you know, in the early days of the United States, books were very expensive and hard to get. The first inventory of the Library of Congress was...
They sent a list of what they wanted to this bookseller in the UK, and it took five months for them to find all the books they wanted and send it to them, along with a bill for like $489. Which, you know, in the 1700s, that was a lot of money. I'm entertained at something as fancy as the Library of Congress, you know, lending out books to Congress people, having to deal with something as petty as library fines. Yeah. It's like humans have always been humans. Even, you know, the founding fathers were going to lose a library book every once in a while. Yep. I left it in my other horse.
Yep. No, it really sort of humanizes them in some way. But yeah, and then the transition from that into what eventually became the catalog that we remember from the 90s was all about solving problems like the ones we've been talking about. Like, okay, we've got a new book and logically it should go between these two, but there's no space on the paper to put it between those two. Or it's very easy to find a book by author here.
but much harder to find one by subject. And those are all the things that led step by step to Library of Congress moving away from one big table to eventually cards to eventually software. It kind of sounds like they stumbled across normalization principles here. They're like, oh, we're gonna put everything in one table and then we run out of space or we're duplicating this information all the time and now it's difficult to change. Yeah, none of our database problems are new. They're just digital. I feel like when we talk about data structures and algorithms and software.
We tend to think of them in terms of some very particular sets of properties that we care about. So things like how easy is it to search or what is our worst time cost to access a random item from the structure? Or maybe we care about space. How much space does this structure take? If we need to insert, how much work is it to insert versus to read versus to remove?
How fast is it to search by different arbitrary attributes? How much of those are concerns when you're thinking about more physical information structures? Are those things that are as central to the discipline of information science as they are to sort of classical data structures and algorithms class? Oh, yeah. I think all of those are relevant, right? We don't want to run out of space. I feel like space is even more of a relevant question when you're talking about physical storage, right? It's a lot easier nowadays to jump your AWS database up to the next tier than it is to make a physical room bigger. Right.
It's harder to auto-scale. Yeah, there's less auto-scaling in the physical paper data structure world. Yeah, and like write speed, huge. Do I need to go when, you know, if this author passes away and we've got information about their biographical details written in five different places, then I have to go find all five of those places to update it. Which, yeah, I feel like it's all of those sort of characteristics are important for physical or paper data structure even more so than for a digital one because the cost of any individual action or any individual storage space.
is higher in a human-maintained physical paper structure than it is in a digital one. That makes sense. So the fact that even the most optimized search for a human going through a card catalog or a paper table... is going to be way slower than a database one. So when we start to do it poorly, the costs of it and the downsides are bigger too. Do information scientists think about things in terms of like big O notation? Or is that something that's sort of truly in the software world? That's a good question. I never encountered big O notation in my information science studies, which doesn't mean nobody does. But it wasn't a big part of what I remember.
But I think there's also, it's hard to draw a distinct line between information science and computer science. Information science is sort of like abstracted computer science, but a lot of the ideas and principles and questions overlap. And that's true for information science and library science also. I've always sort of thought of it as like the connector between library science and computer science is information science, sort of an abstracted version of both. That makes sense. Even sort of more university level computer science tends to be a little bit more abstracted. Yeah. Right. You're not talking about Ruby's particular implementation of an array on a 64-bit machine or whatever. You're talking about in a sort of world where you have a data structure that are consecutive chunks of memory, but very sort of that memory could be a paper system. It could be a digital system. It could be. What are the mathematical properties of accessing that?
Yeah, and I think that becomes really obvious the farther back you go in the world of programming when, you know, punch cards, that's a physical piece of paper that you're using to organize your code. Right. Some classic computer science was sort of done before we even had the machines to run them on. Right. So when Turing posits the Turing machine, it's a thought experiment. Imagine we had a machine with these properties. What would the math on that look like? And then it turns out eventually we did build a machine that had those properties and that math turned out to be really useful. But the thinking about the abstract limitations of computation in many cases predate the physical ability to do so.
Yeah. So I guess the answer is that paper data structures and digital data structures are exactly the same, but the digital ones are just faster. Yeah. And I guess they keep getting faster as our hardware gets better. Yeah, they do. Do you feel that studying paper data structures and more broadly thinking about organizing information has made you better as a software developer? I think so. I think the thing that was the perspective from my time with information science.
is that it was kind of inherently user-focused, right? Because we're not discussing things related to a specific medium, right? Like we're not talking specifically about software or specifically about libraries or specifically about paper. We're talking about just generally accessing information. And so the main thing that you have to think about there is like the user, the person who is trying to access this information, what do they want? What information do they have that they can put in? And what do they expect to get out? And so keeping that in mind when designing software, I think makes it easier to think about like, okay, yeah, we could spend all day calculating.
the O notation of this algorithm, but like, is it even giving the user what they actually want? So I think my time studying computer science has helped me think more abstractly about problems, but also more user-centered about problems. I love that dualism. Yeah, it's less about software and more about, here's a problem that we're solving. We happen to be using software to do it. And I think that right there is probably the most important takeaway. Yeah. On that note, shall we wrap up? Yeah, sounds good.