[jjnoakes]

In the first pass, 6 bugs were found and reported. 3 heap-use-after-free, 3 heap-buffer-overflow. Similar numbers in the second.

I'm so glad new programming languages are making strides which prevent this sort of thing outright. They don't prevent all bugs, but they sure prevent some of the most damaging ones.

[theonemind]

This seems like a fairly dubious claim without something to back it. What makes implementations of newer languages more immune to heap-use-after-free and heap-buffer-overflow bugs?

My first guess is that the two main things causing this is number of lines of code (attack surface), and being implemented in C. The reference implementation of Python is also in C. GHC is partially implemented in C. Being implemented in C still seems common, and C is prone to these kind of bugs. Perl is a complex language, giving the code base a large attack surface. If anything, I'd guess languages with smaller implementations have fewer bugs, since one of the most robust findings on these sorts of matters is that the number of bugs is proportional to the number of lines of code independently of other factors.

[viraptor]

> What makes implementations of newer languages more immune to heap-use-after-free and heap-buffer-overflow bugs?

Simply the enforced restrictions on memory management. Of course some of those languages are implemented either in C, or in some language which implements that memory management. But if you implement a language with forced GC and bound checks (python for example), you have only two places where use-after-free and overflows can originate. (ignoring native extensions) It's much simpler to review / fix them globally rather than in each usage site. Putting other restrictions like strict type hierarchy, lifetimes, escape analysis, etc. in place makes the languages more immune by design.

Older/newer split doesn't make a huge amount of sense though. There are lisps, there's ADA, there are lots of other very old languages which avoid those issues by design.

[SwellJoe]

I believe it'll be interesting to see how things shake out when we start seeing languages implemented in Rust. We already have a wide variety of languages for the JVM, which I believe most everyone would agree is a safer environment for a language than C, and I'd bet the theory above holds up. I don't know enough to know how to test that theory, however.

Basically, I think the "newer languages" comment refers to the implementation language (C being the most common today, but probably not for more than a few more years) rather than the language itself (Perl in this case). Though I think you're entirely right that surface area plays a big role. I would be willing to bet that Perl 6 will be more secure (by some definition of "more" and "secure") than Perl 5, because the surface area of Perl 6 that is in C is much smaller (the VM, with most of the language itself being written in a subset of Perl called NQP, or Not Quite Perl).

[rurban]

Not because the surface area is smaller in perl6, only because with GC those use-after-free bugs do not happen. Only with refcounted VM's, where it's easy to access a value with refcount 0. With a GC other memory bugs are common: Forgetting write-barriers, forgetting roots, external data.

And since most of the old dynamic languages (and also the new conservatives ones) are refcounted, the most common error is use-after-free, and then buffer overflows (heap or stack).

off-by-one is common to all systems.

Since I know perl5 and perl6 internals inside out I wouldn't bet that perl6 is more secure than perl5 at all. The perl5 is more hairy and eroded over time, but the perl6 vm is not really battle-proof and way too slow to be serious.

[eCa]

This is Perl5, not Perl6.

[SwellJoe]

I believe previous comment is referring to the implementation language, not the language. Since these are all errors in the C code, rather than errors in a Perl program.