Speculating about what made it stand apart: it seems like the (enforced) simplicity of the interfaces between pieces of code. Just text streams; one in, one out, nothing more nothing less. No coming up with a host of parallel, named streams that all have their own behaviors that need to be documented. And good luck coming up with a complicated data protocol built atop your text stream; it won't work with anything else and so nobody will use your program.
Interface complexity was harshly discouraged just by the facts-on-the-ground of the ecosystem.
Compare that with the average library interface, or framework, or domain-specific language, or REST API, etc. etc. and it becomes obvious why integrating any of those things is more like performing surgery.
Copy & paste is the basic, ubiquitous, doesn't-try-to-do-too-much IPC mechanism that allows normal users to shovel data from one program into another. As simple as it is, it's indispensable, and it's difficult to imagine trying to use a computer without this feature.
The same applies to pipes, even though they work a little bit differently and are useful in slightly different situations. They're the "I just need to do this one thing" IPC mechanism for slightly more technical users.
Remember the first iPhone? I wasn't into it, but a bunch of my (senior developer) colleagues were. I asked them how they lived without copy & paste and they all told me it was just no big deal.
I would guess the reason you dont do it much on iphone is the lack of multitasking support.
Android alt tabbing has always been a basic part of life and now split screening is a thing too.
You lot getting widgets soon, maybe you'll get alt tab next year?
I understand it's a "hidden gesture" and thus all the UX complaints apply, but from pure observation most android users use the app switcher about as often as PC users use alt+tab... basically not as many as you think.
Most mobile phone users who are even slight technophobes will switch apps the old way from the home screen or app drawer. at least with the iPhone the "going home" behavior can accidentally surface the app switcher which after enough exposure might train such users to use it.
Idk, I kinda hate all the UX paradigms that android introduced, it often felt like they ignored much of what xerox parc taught as objectively good UX and tried to go it on their own, only to have to relearn all the same mistakes that Mac and Windows learned over the last 3 decades.
(ignoring everyone young enough to not remember that iphone first many years didn't have copy paste)
The iphone NEVER replaced a computer.
It replaced the television. No need for copy paste there.
The ipad is today trying, very shy, to hint at replacing a couple small use cases of the computer.
> Are we all not glad we don’t use the Unix method of communicating on the web? Right? Any arbitrary command string can be the argument list for your program, and any arbitrary set of characters can come out the other end. Let’s all write parsers.
The way that I think about this is that the Unix philosophy, which this behavior is undoubtedly representative of, is at one end of a spectrum, with something like strict typing at the other end. Rich, being a big proponent of what is described in the article as "Lego-like" development clearly does not prefer either end of the spectrum, but something in-between. In my opinion as well, the future of software development is somewhere in the middle of this spectrum, although exactly where the line should be drawn is a matter of trade-offs, not absolute best and worst. My estimation is that seasoned developers who have worked in many languages and in a variety of circumstances have all internalized this.
And yet, at least for the Unix tools I've used, nobody did write an elaborate parser. Instead, they all ended up using newlines to represent a sequence. There were never really nested structures at all. That least-common-denominator format they were forced into ended up making input and output really easy to quickly understand.
Maybe the problem with Rest APIs is that JSON does too good of a job at making it easy to represent complex structures? Maybe we'd be better off using CSV as our data format for everything.
Say Unix tools returned JSON. What good is that without a schema? Instead of a parser, let's all write schema transforms?
sed and awk are the elaborate parsers that are used in complex unix command line scripting.
The biggest mistake in the design of Linux is that they allow newlines in filenames. Seriously, what the heck?!
Um, but that's exactly what we do use on the web. Oh, sure, there's some really popular formats for input and output with widely available parsers (HTML, XML, JSON), and HTTP itself (as well as other protocols) specifies headers and other formats that need to be parsed before you get to those body formats, including telling you what those headers tell you about the other formats so you can no which parsers you need to write, or use if you can find one written for you.
A lot of the value of scripting, pipes and associated concepts is low barrier to entry.
I agree with the value of strong typing. But I also remember how infuriating it was to learn to work with a type system when I was learning to write code.
When I need a quick answer of some sort, iteratively piping crap at a series of "| grep | awk" is exactly what the doctor ordered. Sure, I could bullet proof it nicely and make it reusable by investing time the time to write it in something saner, but there's zero reason to - I'm not likely to ever want to perform the same action again.
Unfortunately this is complexity in and of itself. I don't disagree that different cases require different tools, however the split should be strongly weighted in one direction. Mostly IPC over pipes with a few exceptions, mostly REST and JSON with a few exceptions, mostly language X with a few exceptions. Everyone will have their own preferences, but I think it's important to pick a side (or at least mostly pick a side) or else you accept chaos
- At the scale of 1,000 to 1M lines of code, authors/programmers have some control, and you're dealing with a time scale of months or years.
- At the scale of 10M to 100M lines of code, no organization is an author with complete control, and you're dealing with a time scale of decades.
In the latter case, you can't make global changes to the system, so you end up with glue in the form of pipes and textual data dumps.
You're going to have some chaos because you can't go and change every system to be consistent.
So I would say the future of software development is at BOTH ends of the spectrum, not somewhere in the middle.
Good material on the subject: https://www.dreamsongs.com/Files/DesignBeyondHumanAbilitiesS...
If you look at the code in a web browser or operating system you will see this "lack of control". Once a codebase reaches a certain size, there's only so much you can do with it, and you end up with loosely coupled glue around it.
I think a reasonably minimal Debian system has to have somewhere between 100M and 1B lines of code in it, and that whole thing becomes a blob that you dump on a virtually free computer like a Raspberry Pi, etc.
I think if you look at the software that banks and airlines run on you will see the same thing. There's some code from the 60's or 70's written in some weird language at the heart of it. You don't really get to architect the system; it's more accurate to say that the system places its constraints on you.
Uzbl is a collection of "web interface tools" that adhere to the Unix philosophy, that come together to create a browser.
He clearly uses a different web to the one I use. >.>
You could almost say that micro-services are trying to follow in the Unix tradition. But the problem is that a) they don't really get used in that ideal, small-scale use-case because they're almost always written and consumed internally, not exposed to the public, and b) they do get used in those huge, complex cases where their lego-ness stops being a virtue and starts being a liability.
SaaS is all the rage now, and the business of SaaS is disincentivized from making "lego-like" software. The easier it is to interoperate, arguably the easier it is to replace or reimplement your service.
Because a lot of the value of unix tools is derived from their interoperability, they play nice with each other.
An interesting inverse-example is Slack, which does lots of interop with other services, because it isn't a trillion-dollar company that can attempt to give you everything you need under a single roof.
Right, just like LEGOS thrive in simple cases, are not suited for building complex systems.
That is why the idea of "Software LEGOs" is not reality. Or you can say that Unix Pipes are "software LEGOs", with similar limitations.
In my practice I have used both with great success. For logging, parsing, and displaying playback data from field systems, UNIX and UNIX-like tools have been incredible. VNLog in particular is a wonderful way to interact with data if you need just a bit of structure on top of unix outputs.
And anyway, getting from not-so-great data to something that a typed library in language X can parse is a great job for plain old Unix tools.
I recently picked up Swift and it was a dream. Never used it in nearly the same degree, but I was highly productive quickly.
The thing is: I don’t think it’s a bad paradigm (everything a stream of text), it’s just the “IDE” sucks completely. If there was inline autocomplete, peek-ahead results, a “verbose mode” where the options are all more like typed full-English objects, instant fuzzy search of actions, multi line with an easy way to edit, etc, it would probably be fantastic and I’d be a 100x shell developer. It would also basically feel like an IDE at that point for a normal programming language!
1) data is a stream of bytes
2) data is only a stream of bytes
That's it. And it turns out that's a pretty powerful abstraction... Except it requires the developer to write a lot of code to massage the data entering and/or leaving the pipe if either end of it thinks "stream of bytes" means something different. In the broad-and-flat space where it's most useful---text manipulation---it works great because every tool agrees what text is (kind of... Pipe some Unicode into something that only understands only ASCII and you're going to have a lousy day). When we get outside that space?
So while, on the one hand, it allows a series of processes to go from a text file to your audio hardware (neat!), on the other hand, it allows you to accidentally pipe /dev/random directly into your audio hardware, which, here's hoping you don't have your headphones turned all the way up.
This example also kind of handwaves something, in that you touched on it directly but called it a feature, not a bug: pipes are almost always the wrong tool if you do want structure. They're too flexible. It's way the wrong API for anything where you cannot afford to have any mistakes, because unless you include something in the pipe chain to sanity-check your data, who knows what comes out the other end?
You used to be able to cat things to /dev/dsp, but- that used something like 8khz, 8-bit, mono audio. That's horrendous. Because, with just a stream of bytes, you have to settle for the least common denominator- /dev/dsp had IOCTLs to set sample rate, number of channels, and bit depth, but... with just a stream of bytes, you can't do that.
Similarly, video data via /dev/fb0 - AFAIK you don't even have defaults there to rely on, to display anything useful you need to do IOCTLs to find out about its format.
When do you not want structure? Seriously-
Plain, human-readable ASCII text is maybe a candidate - but even then there's implicit structure (things like handling CR/LF, tabs...)
Unicode text? You know that's structure. (Ever had a read() call return in the middle of a UTF-8 multi-byte sequence?
CSV? That's structure.
Tab-separated columns? That's structure.
Fixed-width columns? Also structure.
You don't get to not have structure. Structure is always there. The question is whether you get to have a spec for your structure, or whether it's just "well, from the output it looks like column 3 here is always the hostname of the server I want, so I'll use 'cut' or 'awk' to extract it". That approach can work in practice, but...
Of course, one can always botch configuring that structure. You never really quite escape the rule of "garbage in, garbage out."
But that's not what an abstraction is! UNIX pipes are maximally low-level, as close to un-abstract as it is possible to get, except perhaps if it used bit-streams instead of byte-streams. It literally cannot get less abstract than that.
UNIX pipes are completely untyped, like a C codebase that uses only void* to pass all data. (Okay, some way of indicating end-of-stream is also needed, but it's still a good analogy.)
> pipes are almost always the wrong tool if you do want structure
Not almost the wrong tool, entirely the wrong tool.
You can't magically shoehorn types back into an untyped system when all the existing components assume untyped streams.
PowerShell's structured and typed object streams is more UNIX than UNIX: https://news.ycombinator.com/item?id=23423650
If this were true, then a typed system couldn't exist on top of machine code and random access memory, because the existing components (i.e. CPU + random access memory) have no knowledge of the types you are speaking of.
I think the assumption is false. Types can be shoehorned onto untyped data.
To put a spin on it, maximally low-level within it's particular domain, perhaps, sitting on top of the transistors below them, sitting on top of quantum physics.
Thinking a bit further, an abstraction appears to be the mapping in both directions between two domains. That being said, the two domains themselves, are independent, and don't appear to be the abstraction.
The verb "abstract" would mean we are mapping from one domain to the other.
I think "homomorphism" is the mathematical term for what I'm describing.
What's my point? We can always go lower. :)
I agree, and it just hit me while reading your comment that the special thing is not just that you can plug any program into any other program. It's that if one program doesn't work cleanly with another, this enforced simplicity means that you can easily modify the output of one program to work with another program. Unix command-line programs aren't always directly composable but they're adaptable in a way that other interfaces aren't.
It's not great for infrastructure, don't get me wrong. This isn't nuts and bolts. It's putty. But often putty is all you need to funnel the flow of information this one time.
This is why functional programming and lisps are such fantastic development environment, because you can use components (functions) that are not very opinionated about what they are acting on.
BTW if I was making UNIX command line today it would use LinkedHashMaps for everything instead of text streams.
cat myData | toolStripsWhitespace > myData
var myData = RestApi->read(myDataRecordID)
var mutatedData = someMutation(myData)
var myData = new SomeObject(RestApi->read(myDataRecordID))
var myData = RestApi->get('/records', id=myDataRecordID)
myData = RestApi->post('/tools/strips-whitespace', myData)
cat myData | stripWhitespace | allToUppercase | removeTheLetterC > myData
var myData RestAPI->get'/records', id=myDataRecordID)
myData = RestAPI->post('/tools/strips-whitespace', myData)
myData = RestAPI->post('/tools/allToUppercase', myData)
myData = RestAPI->post('/tools/removeTheLetterC', myData)
Of course in the real world, you can put whatever endpoints you want into your API, and there are dozens of wildly different APIs being called RESTful, but regardless of any of that, rarely are they conducive to being used in pipes, because the point the GP was getting at is that tools are made in the UNIX ecosystem with the pipeline in mind, and that is not the case with REST endpoints.
Thanks for the discussion by the way!
I do think that there is an interesting point around "designed with pipelines in mind" though. I don't know if that would specifically make things more "lego-like", or if that would even specifically be a good thing. But it would be certainly interesting to analyze the list of "average library interface, or framework, or domain-specific language, or REST API" and evaluate each of them along those lines, and consider what it would look like if each was designed them to be more pipeline-centric.
Or maybe the original post came more from a feeling of "it's easy for me to think in terms of pipelines, so anything that forces me to do otherwise feels needlessly complicated"?
But, if we start thinking in terms of pipelines, then that's essentially what functional programming is, run a function, get this output, pass it to another function, get its output, run another function, get its output. So, maybe the love for pipelining comes out of actually a desire for a functional core. So it'd be more like "i only want to think in terms of functions at the core and not have to worry about side effects, and unix does this on a program level".
So, of course, you can't entirely get away from side effects, otherwise you'd literally be doing nothing. So we push it to the edges as far as we can. The analogy for side effects for Unix then would be `> myData` as the last step. So you could say that Unix's philosophy is to have programs that form a functional core, with there being only one possible side effect, which then only has one possible placement, which is at the end of everything. Sounds like a functional dream world.
Back to the list, if we wanted to design "average library interface" with this idea, then maybe you'd say here is a package with all the functional methods. And here is a package with all the side effect methods. And the design is to run a bunch of methods from the functional package before finally making a call to the side effect package.
If we then loop all the way back to REST APIs (and I mean REST spec this time) then I think I could make an argument that REST does in fact follow this functional core philosophy. You have GET endpoints which are your functions, that let you retrieve resources. And then you can POST against objects, which are your side effects, which let you create or update a specific resource. The spec limits the types of side effects you can make, almost to the point of making them dead simple. So, we then have a clear delineation between what is a function and what is a side effect, with a very limited and clear number of side effects that you're able to create.
If you ever wondered why some people are obsessed with functional programming this is the reason why:
Functional programming forces every primitive in your program to be a Lego Block.
A lot of functional programmers don't see the big picture. They see a sort of elegance with the functional style, they like the immutability but they can't explain the practical significance to the uninitiated.
Functional Programming is the answer to the question that has plagued me as a programmer for years. How do I organize my program in such a way that it becomes endlessly re-useable from a practical standpoint?
Functional programming transforms organ transplantation into lego building blocks.
The "lego" is the "function" and "connecting two lego blocks" is "function composition".
In short, another name for "Point free style" programming is "Lego building block style" programming.
There's no reason you can't do this with functional programming, but obviously you can do it with non-functional programming too, and you could certainly fail to do this with functional programming.
You're not free from the need to specify how to parse the strings just because your whole human readable system is stringly typed.
It comes with the same drawbacks (the lack of static typing is similar to strings with Unix; you have to be aware of the details of the data coming in and out because there is no type system to save you - unlike Haskell), but the benefit is easier composition.
In either case you have to compose functions With the correct types otherwise you hit an error. The difference with dynamic types and static types is when this error occurs. Run time for dynamic types and compilation time for static types. Either way some type of error will always occur regardless.
Though I agree with you if you have a limited number of types things become easier to deal with.
But being able to specify interfaces for lambdas is absolutely fantastic. Something like: "this function takes an object and returns a function that can take either a string or a number and returns a number" is easy to do in Typescript, and useful for figuring out where the lego block fits.
I.e. this will warn (inc "hello") in clj-kondo
Good interfaces enable unexpected things.
Obviously not, the types have to be compatible just like how a lego piece must be compatible.
>The programmer has to have the foresight to make their types compatible.
Just like a lego builder needs to have the foresight to see whether two lego components are compatible. The analogy still fits.
> I think the novelty of the Unix pipe for interoperability is that they (typically) work on one agreed-upon kind of data: human readable, white-space separated. So a lot of tools "just work" with each other.
The agreed upon data is just a type. Unix pipes compose functions that take in strings as input and deliver strings as output.
The spaces is just an oversight, the whole thing would have been more clean if the type was designed to represent multiple pieces of data like a tuple of words rather than a single string with words divided by spaces.
>you could certainly fail to do this with functional programming.
Just like how two lego pieces can't compose if they're not compatible pieces facing the right direction. Sockets and plugs must be compatible. A functional program still fits the analogy of legos perfectly.
Take a look at lego pieces here: https://brickarchitect.com/2019/2019-most-common-lego-parts/
There exists lego parts that can never compose with certain other parts. Just like functions and function composition.
What you're not seeing is how other styles of programming such as OOP or procedural programming fail to fit the analogy and literally become like organ transplantation.
Let's take a look at object composition. How does that work in the context of legos? It's like a lego block with a mutating hole in it and once you can put different things in that hole and that changes the overall lego block.
Also with OOP and procedural programming you get lego blocks that can mutate. Lego blocks can transform themselves and other lego blocks. They form an interconnected network of entities that are constantly mutating.
To fit the analogy of a lego block you need unchanging blocks.
The entire field of Procedural programming and OOP is basically the art of building systems with mutating primitives that constantly change. Imagine constructing buildings using bricks that change shape. Now imagine refactoring code that does this.... Organ transplantation.
Also you really just need to try it. Doing the type of grafting and refactoring that's an intrinsic part of programming literally becomes lego-like once you program using the functional style. To really see the big picture I suggest you exclusively try the point free style using Haskell.
That being said. There are still massive tradeoffs to doing functional programming. But if your goal is to program like you're using lego blocks, functional programming is the path.
That kind of sounds like a structure of living cells. They're more complex than a structure of lego blocks, but they can do a lot more too.
Don't take that as an argument in favor of OOP though. Building with legos is a lot easier to get right than building with cells would be.
The key lies in your choice of primitives. Do you start off with a set primitives that are extremely complex or a set primitives that are simple but complete in the sense that you can build entities of unlimited complexity by composing simpler primitives?
I would say the latter method is easier. But there's no reason why the former method won't work. Biological evolution has built many examples of working machines with primitives of unimaginable complexity.
Basically, it's the infinite monkey theorem.
Pure functional programming is complex, one has to compose effectful functions differently to pure functions. But the pieces do really fit like Lego bricks, especially when the same mathematical abstractions are used consistently. The Haskell community has been extremely effective at creating such consistency, by promoting various abstractions using category theory as a guide. The object-oriented community is not so different in this regard with their promotion of "patterns".
I've been writing Haskell professionally for 8 years. The problems with Haskell are the tooling, language stability, the learning curve and the difficulty of reasoning about performance, especially space usage. But composition and re-use works.
The key ingredient in making software reusable and portable is the talent and experience of the engineer, regardless of language.
If it's an abstract data type, it doesn't have to be understood by the caller, it just has to be passed to another "lego brick" which understands the abstract interface. If it's a structured data-type, then there's no reason why it shouldn't be understood by the caller, the type should tell you how to consume it. I do not understand your point.
> So the solution is to make that data easily convertible or generic.. and guess what? That's no different than doing the same in an OO language
I guess you mean structured data types here? But you have missed my point completely about side-effects, it is side-effects (coupling via back-channels) that prevent composition, in general, in an OO language. OO languages also typically have an obsession with nominal types that can impede reuse.
> The key ingredient in making software reusable and portable is the talent and experience of the engineer, regardless of language.
You appear to be suggesting that languages/tools don't matter? Why not aspire towards languages that encourage safe composition and re-usable software? Your argument reduces down to "good motorcyclists don't need helmets".
> If it's an abstract data type, it doesn't have to be understood by the caller, it just has to be passed to another "lego brick" which understands the abstract interface. If it's a structured data-type, then there's no reason why it shouldn't be understood by the caller, the type should tell you how to consume it.
And this is exactly how any well constructed OO API works as well.
You can have really good and really bad APIs in any language, it really is up to the skill of the developers. I firmly believe that, there’s no language or tool that suddenly makes you create better things.
I have in my jobs interacted with high profile robust APIs in C++ as well as functional languages. They can be a joy to use when designed well, regardless of language.
Well it depends on the type of composition you're talking about. Anything when mashed together hard enough can compose.
I recommend you try building something complex with the point free style using Haskell. It will be a bit mind bending at first but if you get it then you'll see how FP is basically the same thing as using legos to build data pipelines... exactly the style used with unix pipes.
Only in the same sense that all computing is Turing Machines or NAND gates.
This is a very common misunderstanding, but there is a reason that FP, which is transformational, had to adopt dataflow in order to sort-of handle reactive systems.
Functions run to completion and return their result. Filters tend to run concurrently and, importantly, do not return their results, they pass them on to the next filter in the pipeline.
It is possible to compose any two filters. It is not possible to compose any two functions, not even close, the default for functions is to not compose (arity, parameter types, return type,...).
Filters are functions. They are one and the same. For unix it's a function that takes in a string and outputs a string. The unix world reduces everything into a singular type. You will note that unix "filters" cannot compose with functions of "other" types either. If unix included types like arrays or ints in stdout you would have the same typing problems as you have with functions.
In the unix world all your types are strings so things seem simpler, but in reality your little unix programs need to deserialize the strings into proper types in order to do anything meaningful. You could achieve the same thing if you used functions that returned strings all the time then did internal deserialization but that's not really a solution.
I do get your point about arity and parameter types. You're referring to the fact that not all functions can compose UNLESS they have compatible types and an arity of 1 in the parameters and an arity of 1 in the return value (Golang for example can have arity > 1 in the return value).
However this arity issue doesn't really exist. A function with Arity of two is isomorphic to a function of arity 1 that takes in a 2-tuple as input.
f(a, b) -> c
f'(Tuple[a, b]) -> c
g(d) -> Tuple[a, b]
The only compatibility problems with composition among functions in the end is just types (not arity) but this problem still exists even in the unix world... it's just hidden from you because you don't actually watch the programs perform deserialization and serialization.
As I explained before: no they are not.
They have some similarities, but they also have the important differences I outlined above.
The single type is also a restriction relative to the functional model, and crucial to (syntactic) composability.
> If unix included types like arrays or ints in stdout ...
But it doesn't...
> you would have the same typing problems as you have with functions.
...so you don't.
> Both f' and f are equivalent in theory.
But they are not actually equal (never mind the same), at least not in typed FP, and most FP is typed.
> it's just hidden from you
Exactly. It is hidden in Unix and not in FP. That's a difference.
> You could achieve the same thing if you used functions that returned strings
(1) you write that you "could" achieve the same thing. Meaning that, once again, they actually are not the same. Otherwise the "could" wouldn't make sense (and it does).
(2) returning strings would not be the same at all, because filters do not "return" their results. They return status codes. The result is not returned, it is passed directly to the next filter, continuously.
So: it is good to recognize that 'X is similar to Y', but that doesn't imply that 'X is just Y'.
Let me address this point so it is more clear. Functions by definition do not need to "run" at all. The underlying implementation of a programming language may "run" a function lazily, eagerly or concurrently but that is not what the Definition of a function is describing.
A function is simply a relation between two sets, one set called a codomain, another set called a domain is deterministic. That is all.
So within unix piping the domain is stdin, the codomain is stdout, the function is the unix program. This paradigm fits the definition of a function and as long as that "function" remains deterministic and compose-able it fits within the paradigm of functional programming BY definition. there is no need to discuss "similarities" here, because it fits the definition therefore it is functional programming. Of course we can get pedantic about certain non-functional use cases but in basically you are fully aware of the generality I'm describing. No point in getting pedantic.
Also keep in mind when you "execute" these functions it does not matter HOW these functions are executing whether concurrently, lazily or eagerly. The only important thing is the relation between the domain and codomain. That is all.
>The single type is also a restriction relative to the functional model, and crucial to (syntactic) composability.
Every unix program must internally deserialize this string type and serialize it again for output. The typing must be handled regardless. It seems like you don't need to deal with it but it's dealt with regardless. You also get inevitable deserialization problems of incorrect types. You do not escape typing in the unix world.
>But they are not actually equal (never mind the same), at least not in typed FP, and most FP is typed.
A relation that takes a string from a domain and returns a string that is from a codomain is a function that is part of functional programming. This is EXACTLY what a unix program does when relating stdin to stdout and therefore it IS a function and it is FP By definition.
You can look at it from another direction. A function that returns an Array does not fit the definition of what a unix program should output to stdout therefore a function IS NOT a unix program. BUT a unix program IS a function.
Formally unix programs are just the set of all functions where the domain and codomains are strings.
>>Both f' and f are equivalent in theory.
>But they are not actually equal (never mind the same), at least not in typed FP, and most FP is typed.
The mathematical term is "isomorphic" meaning you get the EXACT same properties going one way vs. the other way. It's the syntactic sugar that prevents composition. Using multi-arity functions and single arity functions with tuples only involves two additional parenthesis:
Arity and tuple types are well described to be isomorphic in category theory. I'm not making this up.
One way to think about an isomorphism is that if two things are isomorphic then they are just different perspectives of the exact same thing. Any differences are superficial or illusory.
>Exactly. It is hidden in Unix and not in FP. That's a difference.
Exactly what? If you're not dealing with the type conversions someone else has too. Just because it's hidden from you doesn't mean it doesn't need to be dealt with. Someone has to serialize things and deserialize it. Even from the command line level you cannot escape types. Let's say you have a unix program that can only take numerical strings from stdin. Let's say you pipe english letters to it... if you do then you will have still triggered a sort of type error despite everything being typed as strings in unix. It's just not handled explicitly as a system error.
You cannot escape typing even in the unix world because it is in the end just FP.
>So: it is good to recognize that 'X is similar to Y', but that doesn't imply that 'X is just Y'.
Except we're not talking about similarity. Like I said above X is just Y BY definition.
Not sure what else I can do to make you understand this.
As such, I am bowing out as I can only repeat what I’ve written.
Here's what I did: I defined what a function is, then I described how a unix program fits the definition of what a function is...
In short a unix program with stdin as domain and stdout as a codomain is a function but a function is not necessarily a unix program.
I think what's actually going on is you didn't read what I wrote very carefully. You sort of just skimmed over it. I can't blame you, it is rather long and detailed...
But if you want to have a meaningful discussion you need to read it and ask questions about things you don't understand.
> > Unix piping is basically functional programming.
> Only in the same sense that all computing is Turing Machines or NAND gates.
Yes, you can map and analogize long enough until you reach a point where you find what you perceive to be equivalences.
However, what you've done at that point is rediscover that all these mechanisms are computational and thus at some level equivalent and transformable into each other, just like yoo can implement all of this with just NAND gates.
You have not shown what you claim. ¯\_(ツ)_/¯
Reusability on the web. Here is where we are going:
Except you litterally use side effects to communicate. Not really FP, that part.
In a pipeline like
seq 100 | number | grep '[a-z]' | sort | tr -d '.' | tr a-z A-Z
every component is a pure function (byte strings -> byte strings) and the pipeline's results are totally deterministic [if you don't change the locale or collation order with environment variables]. No component of the pipeline makes any change to the filesystem¹.
¹ although both stat(2) and inotify(7) mechanisms could allow other processes to detect the interactions with the filesystem that are caused by running the pipeline, among other process-related mechanisms that system as a whole could use to detect how many times the pipeline has been run, so Unix itself is definitely not side-effect-free for almost any operation
Or any other state change.
Because it's not stateless.
Hence not FP.
Clearly not all command lines you can type in the shell are stateless (reading the disk, network, stdin, clock, or system RNG, among other things, isn't, and none of those things are rare or hard to do) and the shell doesn't specifically encourage you to write stateless code and it doesn't have a formal way to verify whether or not a particular command like is stateless or not. All of these properties are pretty similar to most popular FP language interpreters except for the last two. I guess it is a relevant distinction that your FP interpreter would typically know (as a matter of static inference) whether an expression is stateless, while your shell would typically not know.
If you want to get pedantic fine, but the majority of use cases is stateless.
Probably most complex shell pipelines follow this pattern in practice, but the first counterexample I thought of was shuf, or sort -R. Also, many pipelines that include any kind of interpolation or variable substitution also don't follow it.
A different insight about reasoning about pipelines might be that the byte string (or ASCII string) type is too weak to catch most kinds of errors, and it's not uncommon for one pipeline component to not, in fact, be a total function with respect to the actual data type that you're trying to capture (I don't know the right terminology for this). This most famously happens when you do regular expression substitutions but your regular expression doesn't actually match the full grammar that you're looking for. Then the pipeline can be incorrect as a whole for some inputs, but earlier and later stages don't notice. It can also happen anywhere that different tools have a different implicit understanding of the relevant grammar or structure.
That connects up with all sorts of other ideas which are actually about type safety and parsing more than FP. For example, PowerShell has tried to take the pipeline concept in a different direction, to the consternation of us Unix purists. Its use of typed objects in this context makes more explicit what the contract between programs in the pipeline is supposed to be. There is also a LANGSEC connection in terms of the risks of informal or underspecified parsers and grammars.
I know I've personally written lots of Unix pipelines that were correct for all the inputs that I personally threw at them, but definitely not correct for every possible input. I like to use ! in vim frequently to shell out to a command line to perform a text editing task, and vim has an associated undo and redo which means that sometimes I'm trying several variants until I find the one that successfully appears to do the edit that I intended. Sometimes I do this at an almost preconscious level, which is really strange in terms of thinking about the concept of the correctness of a program (in this case, where the program is literally only going to be used once, with the programmer looking over its shoulder).
On the other hand, if you're trying to build something large (a house, say - a real house, not a toy one), you don't want the lego interface. Sure, you can plug anything to anything, but not all of those connections make sense. Also, for building something that large, legos are too small a building block to be convenient. You want some larger things - beams and sheets of wood and particleboard, pipes, air ducts. You don't want to have to build all of those out of tiny blocks.
In the same way, I wonder if building larger applications out of FP is going to be similarly tedious. I have never done it, so I will admit that I don't know.
Here, he had a thought, and he expressed it in two paragraphs. I'm sure he could have riffed on the core idea for another 10 paragraphs, developed a few tangential lines of thought, inserted pull quotes -- in short, turned it into a full-blown essay.
Given that his blog serves, at least in part, as an advertisement for his services, he even has some incentive to demonstrate how comprehensive and "smart" he can be.
His unpadded style means I'm never afraid to check out a link to one of his posts on HN. Whereas I will often forego clicking on links to Medium, or the Atlantic, or wherever, until I have looked at a few comments to see whether it will be worth my time.
This is my preferred summary of Strunk and White.
I haven't thought about this very much, and there is a lot I'm curious about that he hasn't elaborated on.
What are the signs of rejection? Whats an example of failure, are there examples of that wonderful modular behavior that he admires?
Its a nice way to introduce a thought or observation, but I want to know more about why he thinks that, not what he thinks.
Not only am I sure it made it easier for him to grade, but it really forced students to write concisely about their work.
You can then project your own opinions into the general framework, and you find you fully agree :)
As a consultant, "I 100% agree with you, you understand me" is exactly the feeling you want.
He writes the long articles that show off his smarts in fairly specialized areas, where you need to be an expert to disagree.
It's really clever, and I'm curious if it's intentional on his part, or just his style.
The teetering edge that grinds some people's gears are monads which don't compose generally but do compose in specific, concrete ways a-la monad transformers. The next breakthrough here, I think, is going to be the work coming out of effect systems based on free(r) monads and delimited continuations. Once the dust settles here I think we'll have a good language for composing side effects as well.
In the current state of things I think heart-surgery is an apt metaphor. The lego brick analogy works for small, delimited domains with denotational semantics. "Workflow" languages and such.
I don’t really think Haskell is _meaningfully_ superior than other languages at the things that OP is talking about.
Refactoring Haskell _in the small_ is much nicer than many other languages, I don’t disagree on that point. Despite this, Haskell applications are _just as susceptible_ to the failures of software architecture that bind components of software together as other languages are.
In some cases I would even suggest that combining two Haskell applications can be _more_ fraught than in other languages, as the language community doesn’t have much in the way of agreed-upon design patterns that provide common idioms that can be used to enmesh them cleanly.
 I’m mostly belaboring these points to establish that I’m not talking out of my ass, and that I’ve at least got some practical experience to back up my points.
 This is to say when one refractors individual functions collections of interlocking abstraction
Glomming together functions that operate on very abstract data structures feels a lot more like Legos than wiring traditional imperative/OO code.
I think it's more complicated than this. Yes, you can push poorly-architected Haskell to production & be in a rough spot. However, my experience says that even the gnarliest Haskell is easier to improve than any other language.
Because of the types, purity, etc, I find that it's much easier to zoom around a codebase without tracing every point in between. I can typically make one small change to "crack things open" , follow GHC's guidance, and then go from there. I've been able to take multiple large Haskell projects that other engineers deemed unfixable (to the point where there were talks of rewrites) & just fix them mechanically and have them live & improve continuously for years to come.
The big thing with Haskell IME is you don't really need to have design patterns that everyone follows. I don't freak out when I see multiple different idioms used in the same codebase because idgaf about folk programming aesthetic. If an idiom is used, I follow it. It's all mechanical. I barely use my brain when coding professionally in Haskell. I save it all for the higher-level work. Wish I could say that about professionally programming in other languages of equal experience :/
So while it's just as susceptible (because good vs bad software architecture is more a function of time & effort) it's also typically pretty braindead to fix.
 A favorite technique is to add a new case to a key datatype and have its body be Void. Then I just follow the pattern match errors & sprinkle in `absurd`. I now have a fork in the road that is actually a knowably a no-op at runtime.
What things make WAI easier than the equivalent abstraction in $other_languages?
(Not going for a "gotcha" tone here, genuinely interested as a person who has tried and failed a few times to write "useful" software in Haskell. Would like to get there one day).
type Application =
(Response -> IO ResponseReceived) ->
This leads to a very nice definition for middleware as the following type alias:
type Middleware =
This means that WAI applications have a nice and easy to understand top-level interface, and that complex chains of WAI middleware can be built up by chaining smaller middlewares together.
The potential benefit here (over other language frameworks) is that the “grammar” being used to describe applications and middleware is the same “grammar” that’s used in most other Haskell applications
(i.e. function composition). Ideally, this should make it more easily understandable to a Haskell practitioner who might not be intimately familiar with the framework at first glance.
I believe you that the in the large it's still hard. It seems so much more pleasant day to day untangling that big ball of string with Haskell rather than Java.
Completely agnostic problems do exist and modules to solve those can be very strong - but that is a small subset of all the problems we want modules for.
In Python, for example, whenever I'm doing operations on lists, I need to consider every function in isolation - will this break for an empty list (zip), or will it break when given a generator (anything that traverses twice)? If I pass a callback with side effects, how will it behave differently if someone calls it twice?
In a a more rigorous ecosystem, I can instead only worry about edge cases on classes of functions/data types - In Haskell, when doing the same operations, I can generally assume that traversal functions were written with consideration for functor/foldable/traversable laws, and those laws tell me how I can safely compose or swap them out without breaking expectations around those edge cases. Higher Kinded Types let code be constrained to work strictly with those abstractions, so I mostly just need to worry about "does this typecheck" and, very rarely, "is this an unlawful traversal". With things like linear and dependent types entering this mix, library authors can communicate even more subtle expectations at compile time instead of leaving users to try to anticipate them or find them at runtime - when you do need to explicitly consider edge cases, the turnaround time on discovering them is important.
However, databases are cost effective to build and distribute, but implementing that level of rigorous proofs for your e-commerce app is not only generally undesired but often increases the difficulty in reaching profitability and can sink a company. Working on system development isn't limited to academic concerns, if you're guiding project development you have to make really hard calls about correctness vs. cost and I say this as someone who likes to be a perfection and do things the right way whenever possible.
So I don't disagree that careful decisions can scale with abstractions - it isn't the case that it is impossible to build a well designed large system, but it will be expensive and at the end of the day we all (probably?) need to balance cost of investment vs. value of that investment - or are having someone higher up the chain making that decision for us.
However the benefits of sticking with those names and jargon is that it only needs to be explained once. Analogies and metaphors help when you're starting out but you don't need to keep them around. It makes speaking precisely with colleagues in the ecosystem easier.
Proofs are on another level. I've also written proofs and they require much more rigor than you will find in Haskell. It can still be done without a degree but I agree -- not useful for most line-of-business applications.
However the power-to-weight ratio of learning the abstractions available in Haskell is smooth. It takes effort to acquire but it pays off in spades. That's where the "scale" comes from.
Polymorphism combined with composition means that once I know a type implements "Traversal" I can use the entire language of traversals with those data types without knowing anything else about the value I'm working with. The more things that implement traversal the more rich my program becomes.
It's not that you can't do composable software with OOP... it's rather that functional programming is all about composition.
See JVM (garbage collection), React, Datomic
Functions and scalar values is probably enough
1. Expansion of the software universe. Back in VB6 times, there were fewer programmers but many languages. Reusing components made with different languages was a big deal (VB used COM/ActiveX machinery to make this possible), but today there are so many more developers, that each language/ecosystem is big enough to exist on an island and happily not interact with anything unless it's a grpc/rest endpoint.
2. Transition to SaaS. We no longer use the same platform for building and running. Your VB app used to run on more or less the same computer it was built on. SaaS applications run in weird, custom-made, hard to reproduce and fragile computers called "environments". They are composed from all kinds of complex bits and this makes SaaS applications less portable. Frankly, they feel more like "environment configuration" than "code" sometimes.
On the other hand .. look at how much software goes into e.g. car entertainment systems. How many systems have "curl" in, for example. Have a look at the vast list of BSD license acknowledgements in many systems.
Look at the npm ecosystem, where people complain that there's too much componentisation.
We don't fully get back to component reuse, but it makes the sharing of services much more feasible, and more portable as well.
So basically, increased complexity of components, any components, makes them harder to combine - except in very carefully controlled circumstances.
So does HTML/CSS/JS, and they'll also often be adapted to the various popular front end frameworks of the day - Angular, React, Vue currently.
I also find it easier to customize and compose 3rd party UI components on the web than I did back in the 90's with VB, Delphi, MFC, COM, etc.
It's neither as bad as an organ transplant, nor as easy as LEGO.
It is also highly variable, dependent upon the SDKs and API choices.
I've written SDKs for decades. Some are super simple, where you just add the dylib or source file, and call a function, and other ones require establishing a context, instantiating one (or more) instances, and setting all kinds of properties.
I think it's amusing, and accurate, in many cases; but, like most things in life, it's not actually that simple.
I try not to go by "Software Developer" anymore because too many managers have gotten the idea that it's something you just do. Which it is if you're only working in frameworks... the stick houses of code.
I consider myself to be an "Engineer," but I forgot all my calculus, many years ago (atrophied). There are those that assume any "Engineer" must know calculus.
I consider myself to be an "Engineer," but I fail miserably at most of the binary tree employment exams. There are those that assume that any "Engineer" must know binary trees (I never came up through a CS curriculum, and, in 35 years of coding, I have never run into one single binary tree).
I consider myself to be an "Engineer," but have never sat for any kind of board exam. There are those that assume that any "Engineer" should have at least one board certificate on their wall.
I consider myself to be an "Engineer," but hold no college degree or certificate even remotely connected to software development. There are those that assume that any "Engineer" should have at least one "Engineering" sheepskin on their wall.
What I do have, however, is 35 years of consistently delivering (as opposed to just "writing") software products. Not all were ones that have made me proud, but I have always been about "ship." I still keep that up, today, even for free, open-source projects.
I have Discipline. Lots of Discipline. I write reasonably efficient, localized, accessible, error-checking, high-quality code, that can be maintained after I move on.
I practice Configuration Management.
I have a consistent coding style.
I have consistent processes, even though it may seem as if I am "shooting from the hip." I write about that here: https://medium.com/chrismarshallny (several articles cover my process).
I try to keep up with current, practical technology without getting sucked into the "buzzword maelstrom."
I practice and support Usability, Localization, and Accessibility in my work.
I take all that extra time to make sure that my code is "ship" code; not just "code." That means lots of really boring stuff that always takes a lot longer than I'd like.
I test my code six ways to Sunday (My testing code usually dwarfs the product code. I write about that here: https://medium.com/chrismarshallny/testing-harness-vs-unit-4...).
I document the living bejeezus out of my code (I write about that here: https://medium.com/chrismarshallny/leaving-a-legacy-1c2ddb0c...).
I have always stood behind my code; taking Responsibility and Accountability. Never lying, and admitting (and fixing) failings, learning from those that went before me; while treating my superiors, peers, and subordinates with the utmost respect.
I have always been keenly aware of the way that my customers used my code, and have always welcomed negative feedback (I write about that here: https://medium.com/chrismarshallny/the-road-most-traveled-by... -scroll down to "Negative Feedback Is Incredibly Valuable").
I am not afraid to admit ignorance, or ask questions. Almost every project I start, I don't know how to do. I write about that here: https://medium.com/chrismarshallny/thats-not-what-ships-are-...
Apparently, my employers also considered me to be an "Engineer," despite these awful failings. I had them fooled, eh?
I've learned the value of speaking from personal experience, as opposed to vague, hand-wavy, often passive-aggressive, "keyboard warrior" stuff.
I find that personal experience, and personal ethos tends to be a lot more relevant to most folks, over personal opinion, especially when it is applied as a projection.
In fact, one of my editing disciplines is to do a search-and-replace for "you" and replace it with "I" or "me."
You'll see a lot of "we," "us," and "our" in my writing, as well.
FUN FACT: Did you know that I nuked a number of passive-aggressive slaps from this posting? I'm a really experienced troll. I have made it a point to change my stripes. One way that I am doing this, is putting my brand and personal information behind everything I post.
I feel like this is trying to argue for more “consulting surgeons” when we need more “tooling machinists” who know how to make a good LEGO block.
True lego have the property that if two parts snap together, it's a valid configuration. That's definitely not the case for unix utilities. You have to read a manual for each sort of brick.
It is rather unfortunate for us that everything that's come afterwards has been in some way even worse.
Obviously shuffling data around is a different (easier) beast, especially for one off tasks.
* interface surface area (i.e. how much of an interface is exposed)
* data types of data coming in and out (static or dynamic). Static languages have an advantage here as many integration constraints can be expressed with types.
* whether it is a very focused functionality (e.g. extracting EXIF from file) vs cross-cutting concerns (e.g. logging)
The more limited surface area, the simpler the data types and invariants, the more localized it is - the more it is like LEGO as opposed to an organ transplant
The reason it works is because unix softwares agree to a simple contract of reading from stdin and writing to stdout, which is very limiting in terms of concurrency but unlocks a huge world of compatibility.
I wonder if we will ever get software legos without the runtime bloat from forking.
ps: to anyone countering with examples of languages that are reusable through modules, that doesn't count because you are locked in to a given language.
In a sense, shared object files / dynamically linked libraries meet this criteria -- they can be loaded into program memory and used by a single process.
There's also flowgraph-based signal processing systems, like gnuradio, which heavily use the concept of pipes (usually a stream of numbers or a stream of vectors of numbers) but, as I understand it, don't require OS forking. (Though they do implement their own schedulers for concurrency, and for gnuradio at least, blocks are typically shipped as source so I'm not sure whether that counts as reusing executables vs. reusing source code.)
Building a system from components can be quite tricky, as the components tend to only play nicely together if they've been built for one another, or to some agreed upon spec. Otherwise you have to be very fluent in all of them, and write glue code. And there be dragons.
APIs tend to be all over the place, yes, but you can have a well defined API spec that really can allow specific implementations to be swapped out fairly trivially, such that it really is more like 'components'. Think SQL, for instance.
I think you may just be noticing that when things adhere to a spec, they tend to be able to combined pretty easily; if not, they don't.
I wrote about it few weeks ago: https://skyalt.com/blog/dsl.html
>I have a plan on how to do that on paper, but because connecting assets together can be complex, it's better If most of the users don't do that. The key thing is that users don't share only assets, but also how and where they are connected to each other. It's like when someone makes and publishes a library, but also example code for how to work with a library. But in SkyAlt, asset connections will be automated. This is also why it's very important to build and grow the community - more assets, more connections, which means easier to use.
Absence of such incentive appears to be the reason that open-source software is not always perfectly connectable -- few people have a significant incentive to ensure this design goal.
I believe that in the future the most of software will be open-source and If people don't have a problem sharing blocks/widgets/assets, they will not have a problem to share information about connections.
Side note: Every program has 2 parts - instructions and data. When people share assets and connections, both are instructions, so their data stay private.
The point of frameworks is to provide the standardization of the "shape" of various bits of software so it's a lot more like snapping together LEGO. But even then, LEGO isn't universally snappable; some blocks just don't click to other blocks in the product line. And then, of course, there's the "illegal" block hacks (https://crafty.diply.com/28429/14-illegal-lego-building-hack...) that work in practice but are not at all using the tool the way it's specified. When software reuse is like LEGO, we should expect (a) some things we want to do, we can't really do without jigsaws and glue and (b) sometimes, people will do things that the software technically allows but no sane person would call "desired" or "intended."
In fact, the LEGO-to-framework analogy works pretty great. And yeah, outside the context of a consistent (and, I'd argue, opinionated) framework, you're about as likely to have two pieces of software interoperate as you are likely to pick two random chunks of matter in the universe, slap them against each other, and have anything useful happen. I just tried it with a brick and this glass of water on my table. Now I have "wet brick in a pile of shattered glass," but I don't know anyone's going to give me a Series-A funding round for that.
I tell you even a secret. Every piece of software has all the code you need to work with it already - just copy/paste stuff. You only need to know where it is. Every piece of software is it's own tutorial how to make that particular piece of software.
I've done this for years. I call my method of working copy/paste method.
And by reading this, I think I know why I piss so many programmers when I say what I do. And how fast I am with it.
John D. Cook's post shines the light once again on the difficulty of writing reusable components.
The software that requires organ transplant just wasn't written to be reused. It is a matter of effort and cost. You cannot just solve your problem and wrap it in a library - you need to take care of all common use cases, define edge case behavior, test, document... It costs a lot more and that price is hard to justify. In the end it's the product that makes money and library is forgotten as long as it does its job. We need to fix the open-source income ecosystem.
Never has it been easier to find existing code, download it and snap it into your project. Npm, nuget, etc, etc.
Also, on the front end there are no shortage of plug and play widget libraries.
Sometimes we need to take a step back and appreciate how things have improved over the years.
Translatable to software, your reusable piece of code might not be so reusable as you think, missing lots of abstractions and having lots of specific code for just one specific consumer. Taking a piece of functionality from one part of your software and using it for something else might harm the original use due to broken API's, introduced complexity latency and failures, etc. Even ending up with having 2 split versions of your reusable component for each use case that you now have to maintain seperately.
However, this is totally accurate when reusing the code from your previous project on your next one. I developed many large gov services, and each one reused bunch of stuff from previous ones like user management, complete areas of functionality etc. Here, the analogy is adequate, artefacts from previous projects creeep in constantly. This feels more like single project evolving (or devolving) over long period of time (years to decade).
Feels so good to write similar software in 1/10th the time because the hard fought ideas are already won.
When you're talking about not being able to reuse code, I think it depends on how you've designed your system.
Design for reusability seems like it needs to be key goal in development in order to achieve software re-use.
The way to make software reuse more like organ transplant than building Legos is to ignore the advice to keep the interfaces simple and precise. Or worse, don't even bother defining an interface, its data types, or error conditions.
I'm curious to hear from those who have worked on large, sophisticated, Spring Framework apps - does it really feel like snapping pieces together?
pg_dump -U postgres db | ssh firstname.lastname@example.org "dd of=db_dump"
And isn't GNU Radio effectively a domain-specific language for plugging signal processing elements together?
With Haskell and Scala's type-system we seem to come closer to universal Lego. Still, like Lego, it can be hard to combine larger scale composites.
This blog is 100% accurate, and nearly everyone took the wrong point away from it.
There is a very good reason we don’t build anything but the most trivial structures from Lego.
Composability and reusability are great, but not at the expense of suitability for purpose.
You could build a house or a car out of Lego, but it would be worse on almost every possible metric than a regular house (unless your primary use case for your house is to take it apart and reconfigure it easily).
The same goes for software. You could build everything from unix pipes, but that rarely happens past the most trivial scale.
In fact, the clue is in the article. Evolution is not very tolerant of inefficient systems; animals haven’t evolved hot swappable organs, they have a complex interconnected system because 99% of the time, it works better.
Another example, a lot of electronics was (and still is) implemented by combining discrete components with well known behaviours (things like https://en.wikipedia.org/wiki/7400-series_integrated_circuit...). At many levels in electronics you have abstractions upon abstractions (transistors -> logical gates -> multiplexers -> ALUs -> CPUs; communication busses, etc etc)
As for electronics, yep, you start with discrete components, but as you said yourself, you quickly end up with specifically designed components for efficiency.
What often happens with software development is that despite all the abstractions that get built while building one piece of software, it often turns out those abstractions can't be just pulled out and reused in another context because often those abstractions are too leaky.
I think the "original sin" stems from how "cheap" it is to write a line of code, and another one, and another one. It's not that good abstractions are never being built during software development; but when they do they usually take a long time to mature and have to be sought purposefully (and they cost way much more to engineer).
What sets software development apart from other forms of engineering is that it's often "ok" to build a house with mud and sticks that will collapse on the first strong gust of wind because ... "because it's just a POC", "let's see if we really have customers before doing it right", "making this too well has an opportunity cost of not doing something else", etc etc
>>What often happens with software development is that despite all the abstractions that get built while building one piece of software, it often turns out those abstractions can't be just pulled out and reused in another context because often those abstractions are too leaky.
My question is whether a non-leaky abstraction is even valuable.
Look at a real world example of non-trivial component reuse: a vehicle engine.
It has a single purpose: convert chemical energy to kinetic energy, and they're really expensive to develop.
But you can't just drop a v6 into a Tesla.
E.g. it's a clear boundary between teams that collaborate on the "car" system, it's something some engineers specialize in that is quite different from "suspension system" (and such knowledge can be carried across engine design instances).
But you're right, when we're talking about LEGO-style composability, as you said, it's about having pieces that fit together and can be rearranged in many ways, and the real world is far messier and rarely you have universal composability; there are instances of "bounded" composability; e.g. the tyres for my car likely fit many other similar vehicles, but not all.
Absolutely! I'm not arguing against modularity at all. I think the concept of discrete subsystems interacting is pretty much universal (organs, components, rooms, streets).
>> there are instances of "bounded" composability; e.g. the tyres for my car likely fit many other similar vehicles, but not all.
Again, completely agree.
What I'm arguing against is uniformity of interface.
You wouldn't connect a tyre to a wheel the same way you would mount an engine to a chassis.
> You wouldn't connect a tyre to a wheel the same way you would mount an engine to a chassis.
Yes, you're right, LEGO is extremely uniform, and no real system is _that_ uniform.
And even a uniform interface doesn't mean any combination makes sense.
Attaching a LEGO brick with wheels on the roof of a LEGO car is not different than soldering an axle on the roof of a car: technically possible but utterly useless.
Actually, soldering an axle on the roof of a car is composing uniform interfaces: atoms in chemical bonds form a finite set of building blocks too!
Now if you instead look at Trucks and Campers, Rear Hitches, Fifth Wheels Hitches, Goose Neck Hitches. Every truck you can buy on the market today (with some strange exceptions Cybertruck cough) has very simple compatibility with all of these various trailer types. Yes in some cases you have to buy a mod kit, but when you look at the mod kit it literally bolts into pre-drilled holes in your truck frame. It's all plug and play.
I feel like the reason it's like this is because no one person can go off and design a truck in the dark. It's designed by many people and is beholden to customers.
Software, on the other hand, just gets built by whoever wants to build something. It doesn't have to fit some kind of standard. And it's not like there are a lot of standards that already exist in whatever it is that the software is doing.
In the real world however, you have Kubota tractors that need to fit a CAT bucket.
I mean just read this page:
Small quirks to it, but yeah, the rest of the world really is Legos, Software is organ transplant.
Funny, but Lego fans almost interpreted "organ transplant" exactly as "snapping Lego blocks".
Have you guys wondered why some people are so fanatic about functional programming?
Functional programming forces your program to be lego blocks by Definition.
A function with no side effects IS A Lego Block.
If I never discovered functional programming I'd be like John D. Cook, blogging about how programming is like organ transplantation.
Functional Programming is the answer to the question that has plagued me as a programmer for years: How do I organize/design my program in such a way that it becomes endlessly re-useable from a practical standpoint?
If this question plagues you as a programmer I urge you to take a closer look at functional programming. Specifically look at functions and function composition.
...follow the types