Time to make C the COBOL of this century

People tend to be obsessed with bleeding edge technology. But those who truly understand know that "bleeding edge" is an anti-pattern and there's a reason it's called that: it can bleed you as well.

If it ain't broken, don't fix it.

Making C go away will require major rewrites of projects that have millions upon millions of hours of development.

Yep. And it'll be done. Yes it'll take a while, but this is what it means for C to be like COBOL (which also still exists). But the more and more it can be marginalized the better we'll all be security-wise.

The rewrite-it-in-rust gang arrives in 3, 2 …

Cattle not pets. They're just computer languages.

"It's too much work, so let's just not do anything and stubbornly stick with a problematic unsafe language that fewer and fewer people are willing to learn"

Only ones who knows how to make proper tech images are the folks at O'Reily publisbing

It's how everyone who's anyone does code reviews!

Getting?

I know COBOL, its pretty easy to pick up honestly. Its just the 40+ years of context and the "WHY"s that they did what they did that is hard.

With how many people are looking for COBOL I really should just make a resume specific to it, put salary * 1.2 and see if any remote positions pop up. Love my job, but with the latest news...

Can I take a a guess? If not, please ignore the next sentence.

Finance companies with a "database" is a gigantic flat file?

Well COBOL hasn't completely gone away... I don't think anyone expects C to become completely extinct; just very legacy.

D missed its chance. Zig is clearly going to be the successor to C, for people who don't want to use Rust.

Concrete technical answer (one of many): imagine you have a list ("array") of 5 numbers, and you try to print the 10th number in the array. A secure language will say "error! it's a list of 5 numbers, there is no 10th one!!". C will instead print some random garbage (whatever happens to be in the part of memory following the 5 element list), or maybe do something even crazier (try searching "nasal demon"), without indicating that anything has gone wrong. There are many other issues like this with C. You end up with programs going completely into the weeds, turning control over to attackers, etc.

Abstract philosophical answer: Secure languages like Ada and (hopefully) Rust are designed to help you ensure the absence of unwanted behaviours, rather than just the presence of wanted ones. If you want behaviour X, the goal of old languages like C was to make sure you could write a program in which X was present. That was a big enough challenge in the old days that language designers stopped once they reached that point. If you don't want behaviour Y (let's say Y is a security attack), it's up to you to just write the program without behaviour Y. 50+ years of experience have shown that to be inhumanly difficult once the program gets complicated, so you really do need help from the language. Accountants invented double-entry bookkeeping 700 years ago for similar sorts of reasons: to keep small errors in complicated systems from sending the system into a nose dive.

Ensuring the absence of behaviours is the classic problem of proving a negative, so there are limits on how thorough the checking can be, and the technical features (like the notorious Rust borrow checker) can be difficult to use. But if you're willing to endure a certain amount of pain and runtime inefficiency (requiring the program to do a little extra work at each operation to make sure the result makes sense, like the example of the 10th element of the 5-element list), you can make programs much safer than you can in C.

Does that help?

Added: Rust is getting some flak because it is pretty new, is still a work in progress, has various unmet goals, etc. It's not fully baked yet but it is getting there (I'm studying it right now). Ada is an older language that is way more mature than Rust, but is more of a pain to use in many ways, so Rust is currently getting more attention.

It's mostly about memory access. Modern languages throw errors if, for example, you try to reference an element of an array that is "outside the bounds" of the array. C does not - it gladly returns whatever memory address is past the end of the array. So the programmer has to check that the index is 0 <= x < array_size whenever they access a an array entry. That's a pain - so they don't.

Its about the type of operations the compiler allow you to do, more or less. Like sharing mutable references, that can be independently changed in a 'hard to keep track of'- manner. Other factors the compiler tries eliminate include buffer overruns and int overflows e.t.c.

Rust for example sometimes makes trivial things a royal pain, see linked lists for example. It also has a gaping microdependency/supply chain attack prone ecosystem, and the compiler interface is also not stable (afaik, caused some issues in linux). There is also no spec.

I have experience of both, and i love both, but C is my fav. Its often trivial to imagine the codegen with C, and there are no shortage of quality compilers. The language is also small enough that implementing a compiler is actually feasible.

It's about memory management.

In programming terms: allocated memory has to have the data in it that you expect in order for your program to work. The unsafe languages do it by manually ensuring it's good and doing so mostly at runtime, or just assume the data is valid and write code that looks valid and have somebody check it before the program runs, or do a mix thereof. In all cases, it require a lot of human intervention and because humans are fallible with different skill levels, this fail quite often.

Safe languages are either built on top of unsafe languages that are battle tested and do lots of runtime checks behind the scenes (interpreted languages like python, ruby, javascript, etc.). Then there are languages that check actions at compile time like Rust. They tell you that the memory you're trying to access can be modified by another part of the code, which might make unexpected changes and that in order to access it, certain conditions have to be met.

In laymans terms: imagine you work at a storage facility (memory) and have to store and retrieve packages. To know where to store and retrieve them, you have a piece of paper with the aisle, shelf, and rack and position on the rack. That's your pointer. To store something, you have to make space on a rack and put the item there, write down the name of the item (variable) and location on a piece of paper (memory address), and keep it on you.

Imagine keeping all of that in order. You have to make sure you don't write down the wrong location (off by one error), remove a piece of paper then it's not valid anymore (dangling reference), remove a piece of paper without removing the item (memory leak), add a piece of paper pointing to something without actually checking what you expect to be there is there and then retrieve it later, and so many other things.
Those are the things unsafe languages allow you to do.

Safe languages either enforce that before doing certain things, you check stuff (runtime checks) or that before you even start doing anything, you plan how you would do, and that plan is checked.

The crazy storage facilities are what most of our world runs on at the moment and there a whole lot of people who love it because it's simple and they know it. "Just tell the intern to get that box there, I made sure it'll be fine. Trust me, I've been doing it this way for years." meanwhile somebody gets the wrong medicine because a piece of paper said another one was supposed to be on the shelf. There are a bunch of people who have thought about ways to improve it, implemented, tested it, and are using it to manage their storage facilities.

Anti Commercial-AI license

Unsafe rust has proven that it can be an effective alternative here, ideal especially when the consumers are also rust.

Literally no one is stopping you