> The industry and the academy have used the term “object-oriented” to mean so many different things.
I think we can safely stick to how IEEE defines OOP: the combination of three main features: 1) encapsulation of data and code 2) inheritance and late binding 3) dynamic object generation (from https://ethw.org/Milestones:Object-Oriented_Programming,_196...).
The article assumes that C++, Java, and Smalltalk implement completely different subsets of OOP features, which is not true at all. Those languages, including Smalltalk (starting with Smalltalk-76), all implement the Simula 67 object model with classes, inheritance and virtual method dispatch. Simula 67 was the first object-oriented programming language (even if the term was only applied ~10 years later in a 1976 MIT publication for the first time, see https://news.ycombinator.com/item?id=36879311). Message passing (the feature the article claims is unique to Smalltalk) is mathematically isomorphic to virtual method dispatch; and also Smalltalk uses method dispatch tables, very similar to C++ and Java.
Regarding Message Passing and Late-binding, I think it's important to take into account that Alan Kay was working on Smalltalk -- a system that was image based; a system where you could change things as it was running. I think that message passing and late-binding are often championed but then sort of fall flat given standard deployment techniques: build & deploy (often to a ephemeral runtime / container).
Smalltalk can use build & deploy. (Image as cache, not archive.)
"At the outset of a project involving two or more programmers: Do assign a member of the team to be the version manager. … The responsibilities of the version manager consist of collecting and cataloging code files submitted by all members of the team, periodically building a new system image incorporating all submitted code files, and releasing the image for use by the team. The version manager stores the current release and all code files for that release in a central place, allowing team members read access, and disallowing write access for anyone except the version manager."
1984 "Smalltalk-80 The Interactive Programming Environment" page 500
Yes, you could also build and deploy a Smalltalk system, but my point is that the “build & deploy” approach (to me) seems antithetical to the message passing and late-binding paradigms. To use another example, it seems like you lose a lot of the benefits of Common Lisp via Slime (hot code reloading) if you deploy your Common Lisp app to a short-lived, ephemeral environment.
I think the biggest mistake was to teach inheritance as a main feature of OOP. I have done some stuff with inheritance but it was very specialized and it would have been fine without inheritance.
There is a good methodological principle stated by a Soviet philosoph Ilyenkov: to understand the nature of a thing build or describe a minimal working model of the thing. So simple that if you remove any single detail it ceases to work. He himself gave an example of radio: to understand what radio is build a minimal radio sender and receiver of three or four details each. Do not start with a household radio unit that comes in a polished box with lights and knobs; it has way too many unrelated details.
This works very well both for concrete things like radio and for more abstract things like math or Marx's notion of private property. This is also the principle employed by religious and mystical parables or the book "A pattern language".
My OO projects were usually in Java with a DB. They all ran afoul of what Martin Fowler calls the Anemic Domain Model. Basically your objects are data-only, so there's no benefit. In addition Spring injection became ubiquitous, and further killed objects with behavior. The only project using a DB and had objects with behavior was an old one that happened to use TopLink as an OR mapping.
> Basically your objects are data-only, so there's no benefit.
This makes me wonder why most of us use Java at all. In your typical web app project, classes just feel like either:
1) Data structures. This I suspect is a result of ORM's not really being ORM's but actually "Structural Relational Mappers".
- or -
2) Namespaces to dump functions. These are your run-of-the-mill "utils" classes or "service" classes, etc.
The more I work in Java, the more I feel friction between the language, its identity(OO beginning to incorporate functional ideas), and how people write in it.
Java was the first popular language to push static analysis for correctness. It was the "if it compiles, it runs" language of its day, what meant that managers could hire a couple of bad developers by mistake and it wouldn't destroy the entire team's productivity.
I'm not sure that position lasted for even 5 years. But it had a very unique and relevant value proposition at the time.
I'd love a deeper dive on how Objects work in NeXTSTEP. From their brochures, they talk about objects being persistent and distributable, that in a workgroup of people everyone can rely on the objects being up to date.
I've always been so curious what the broader technical ecosystem looks like here. Presumably there are still processes running on systems. But these processes have lots of objects in them? And the objects are using Mach message passing to converse with other processes elsewhere? Within an application, are objects communicating across Mach too?
This is a fun work. It feels like the brief outline for a Speaking for the Dead for OOP. Huge amount of things to lots of different people over time.
Seconding @rawgabbit's recommendation for Casey Muratori's The Big OOPs: Anatomy of a Thirty-five-year Mistake, which really is hunting back and back for the cosmogenesis of objects, and covers so much terrain. Objectogenesis?
https://youtu.be/wo84LFzx5nI
On classes, I get it... tbf though I'm fine with prototype inheritance as well, there's positives and negatives to both approaches... not to mention, there are benefits to not really having either and just having objects you can interrogate or even that are statically assigned at creation (structs).
What's funny on the Method Syntax for me, is that I actually don't like mixing classes that hold data and classes that do things more often than not. I mean, I get the concepts, but I just don't generally like the approach. The only exception might be a controller with a handle to a model(state) and the view... but even then, the data itself (model) is kind of separated as a reference, and don't tend to attach too many variants of state to anything... I'm generally a fan of the single state tree approach (often used for games, and famously via Redux).
On information hiding... I'm generally not too much of a fan of hiding members of an object used to hold data... I mean, I can see filters when you're passing something to the edge of a system, like a hashed password on a user object exposed via an api. But internally, I'd almost rather see immutability as a first class over locking bits and pieces down, then exposing member methods to mutate the object internally. Just my own take.
On Encapsulation, like above... I'm more on the side of the Data oriented design approach. To me this is where you have API surfaces and like above I tend to separate modules/classes that do things, from templates/models/classes that hold data.
I'm mixed on Interfaces.. they're definitely useful for plugin systems or when you have multiple distinct implementations of a thing... but after a couple decades of C#, they're definitely overrated and overused.
No strong opinions on Late Binding pr Dynamic Dispatch... other than I do appreciate it at times in dynamic language environments (JS).
Inheritance and SubTyping imo are, similar to Interfaces, somewhat overrated... I just try to avoid them more than use them. There are exceptions, I'm actively using this in a project right now, but more often than not, it just adds undue complexity. With prototype based inheritance, it's also possible to really slow down certain processes unintentionally.
Strong proponent of Message Passing approaches... it often simplifies a solution in terms of the surface you need to be aware of at a given point. Allows you to construct decision trees and pipelines of simpler functions.
Interesting overall... but still not a fan of some of the excesses in OOP usage in practice that I've had to deal with. I just prefer to break problems up slightly differently... sometimes blurring clear lines of separation to have a simpler whole, sometimes just drawing the lines differently because they make more sense to me to break up for a given use case.
> The industry and the academy have used the term “object-oriented” to mean so many different things.
I think we can safely stick to how IEEE defines OOP: the combination of three main features: 1) encapsulation of data and code 2) inheritance and late binding 3) dynamic object generation (from https://ethw.org/Milestones:Object-Oriented_Programming,_196...).
The article assumes that C++, Java, and Smalltalk implement completely different subsets of OOP features, which is not true at all. Those languages, including Smalltalk (starting with Smalltalk-76), all implement the Simula 67 object model with classes, inheritance and virtual method dispatch. Simula 67 was the first object-oriented programming language (even if the term was only applied ~10 years later in a 1976 MIT publication for the first time, see https://news.ycombinator.com/item?id=36879311). Message passing (the feature the article claims is unique to Smalltalk) is mathematically isomorphic to virtual method dispatch; and also Smalltalk uses method dispatch tables, very similar to C++ and Java.
Inheritance is just the unnecessary coupling of composition and polymorphism.
Regarding Message Passing and Late-binding, I think it's important to take into account that Alan Kay was working on Smalltalk -- a system that was image based; a system where you could change things as it was running. I think that message passing and late-binding are often championed but then sort of fall flat given standard deployment techniques: build & deploy (often to a ephemeral runtime / container).
Smalltalk can use build & deploy. (Image as cache, not archive.)
"At the outset of a project involving two or more programmers: Do assign a member of the team to be the version manager. … The responsibilities of the version manager consist of collecting and cataloging code files submitted by all members of the team, periodically building a new system image incorporating all submitted code files, and releasing the image for use by the team. The version manager stores the current release and all code files for that release in a central place, allowing team members read access, and disallowing write access for anyone except the version manager."
1984 "Smalltalk-80 The Interactive Programming Environment" page 500
Yes, you could also build and deploy a Smalltalk system, but my point is that the “build & deploy” approach (to me) seems antithetical to the message passing and late-binding paradigms. To use another example, it seems like you lose a lot of the benefits of Common Lisp via Slime (hot code reloading) if you deploy your Common Lisp app to a short-lived, ephemeral environment.
Muratori traced the history of OOP to the original documents. Skip to the 1:18 mark if you want to skip to his findings.
https://youtu.be/wo84LFzx5nI
I always considered an "object" to be data with identity and state.
All the other stuff, like polymorphism, encapsulation, etc., I consider "addons."
I think the biggest mistake was to teach inheritance as a main feature of OOP. I have done some stuff with inheritance but it was very specialized and it would have been fine without inheritance.
Back in the day, I used to do OOP with C.
It was a common pattern, back then. We’d pass around structs, and have a small library of functions that accessed/modified the data in these structs.
If you wanted, you could add function pointers to the structs. You could add “polymorphism,” by overwriting these pointers, but it was messy.
That said, inheritance can be very useful, in some cases, like improving DRY. I don’t like to take absolute stances, so much, these days.
There is a good methodological principle stated by a Soviet philosoph Ilyenkov: to understand the nature of a thing build or describe a minimal working model of the thing. So simple that if you remove any single detail it ceases to work. He himself gave an example of radio: to understand what radio is build a minimal radio sender and receiver of three or four details each. Do not start with a household radio unit that comes in a polished box with lights and knobs; it has way too many unrelated details.
This works very well both for concrete things like radio and for more abstract things like math or Marx's notion of private property. This is also the principle employed by religious and mystical parables or the book "A pattern language".
My OO projects were usually in Java with a DB. They all ran afoul of what Martin Fowler calls the Anemic Domain Model. Basically your objects are data-only, so there's no benefit. In addition Spring injection became ubiquitous, and further killed objects with behavior. The only project using a DB and had objects with behavior was an old one that happened to use TopLink as an OR mapping.
> Basically your objects are data-only, so there's no benefit.
This makes me wonder why most of us use Java at all. In your typical web app project, classes just feel like either:
1) Data structures. This I suspect is a result of ORM's not really being ORM's but actually "Structural Relational Mappers".
- or -
2) Namespaces to dump functions. These are your run-of-the-mill "utils" classes or "service" classes, etc.
The more I work in Java, the more I feel friction between the language, its identity(OO beginning to incorporate functional ideas), and how people write in it.
> why most of us use Java at all
Java was the first popular language to push static analysis for correctness. It was the "if it compiles, it runs" language of its day, what meant that managers could hire a couple of bad developers by mistake and it wouldn't destroy the entire team's productivity.
I'm not sure that position lasted for even 5 years. But it had a very unique and relevant value proposition at the time.
I'd love a deeper dive on how Objects work in NeXTSTEP. From their brochures, they talk about objects being persistent and distributable, that in a workgroup of people everyone can rely on the objects being up to date.
I've always been so curious what the broader technical ecosystem looks like here. Presumably there are still processes running on systems. But these processes have lots of objects in them? And the objects are using Mach message passing to converse with other processes elsewhere? Within an application, are objects communicating across Mach too?
There's so much high level rhetoric about. Such as this bit. But I'd love a real technical view at what was happening, what objects really were here. https://computerhistory.org/blog/the-deep-history-of-your-ap... https://news.ycombinator.com/item?id=42111938
This is a fun work. It feels like the brief outline for a Speaking for the Dead for OOP. Huge amount of things to lots of different people over time.
Seconding @rawgabbit's recommendation for Casey Muratori's The Big OOPs: Anatomy of a Thirty-five-year Mistake, which really is hunting back and back for the cosmogenesis of objects, and covers so much terrain. Objectogenesis? https://youtu.be/wo84LFzx5nI
Commenting while reading...
On classes, I get it... tbf though I'm fine with prototype inheritance as well, there's positives and negatives to both approaches... not to mention, there are benefits to not really having either and just having objects you can interrogate or even that are statically assigned at creation (structs).
What's funny on the Method Syntax for me, is that I actually don't like mixing classes that hold data and classes that do things more often than not. I mean, I get the concepts, but I just don't generally like the approach. The only exception might be a controller with a handle to a model(state) and the view... but even then, the data itself (model) is kind of separated as a reference, and don't tend to attach too many variants of state to anything... I'm generally a fan of the single state tree approach (often used for games, and famously via Redux).
On information hiding... I'm generally not too much of a fan of hiding members of an object used to hold data... I mean, I can see filters when you're passing something to the edge of a system, like a hashed password on a user object exposed via an api. But internally, I'd almost rather see immutability as a first class over locking bits and pieces down, then exposing member methods to mutate the object internally. Just my own take.
On Encapsulation, like above... I'm more on the side of the Data oriented design approach. To me this is where you have API surfaces and like above I tend to separate modules/classes that do things, from templates/models/classes that hold data.
I'm mixed on Interfaces.. they're definitely useful for plugin systems or when you have multiple distinct implementations of a thing... but after a couple decades of C#, they're definitely overrated and overused.
No strong opinions on Late Binding pr Dynamic Dispatch... other than I do appreciate it at times in dynamic language environments (JS).
Inheritance and SubTyping imo are, similar to Interfaces, somewhat overrated... I just try to avoid them more than use them. There are exceptions, I'm actively using this in a project right now, but more often than not, it just adds undue complexity. With prototype based inheritance, it's also possible to really slow down certain processes unintentionally.
Strong proponent of Message Passing approaches... it often simplifies a solution in terms of the surface you need to be aware of at a given point. Allows you to construct decision trees and pipelines of simpler functions.
Interesting overall... but still not a fan of some of the excesses in OOP usage in practice that I've had to deal with. I just prefer to break problems up slightly differently... sometimes blurring clear lines of separation to have a simpler whole, sometimes just drawing the lines differently because they make more sense to me to break up for a given use case.
I expected this to be a play on the old joke about Java being designed to appeal to people who were into SM/B&D.