Buffer overflows. That's far from perfect.
It's much too easy to produce buffer overflows in C, be it due to the design of C strings or just because of the manual handling of dynamic memory.
It's much too easy to produce buffer overflows in C, be it due to the design of C strings or just because of the manual handling of dynamic memory.
The big difference was that those OSes versus UNIX costed real money.
Any idea why there was no popular safe array or safe string library for C? Maybe at that time there was no internet and everyone had to create his own abstractions?
Bell Labs was forbidden to sell their research, so they offered UNIX for a symbolic price to universities (vs what other OS used to cost), alongside source code tapes and a liberal license.
This gave birth to several startups that tried to create a business using UNIX instead of the alternatives, given the authors experience with UNIX at the university, e.g. Sun and SGI among others.
Later when the US goverment dictated AT&T split, AT&T was allowed to charge for UNIX and that is when they decided to go after BSD, because 10 years later, under such conditions, UNIX was slowing eating mainframes, given the success of Sun, SGI, Aix, ....
There are no safe array or safe string libraries, because they all fall appart under scrutiny, given C's approach to secure code.
I do not understand, maybe give an example.
About the history I am not sure that proves that UNIX and C advantage was only because of that, it could be a factor but there are merits to C and UNIX that if you want to disprove you can;t just do it by mentioning that history. I mean Windows API used C.
Because since those libraries are not built-in types, you always need to convert back to a ptr and length values at some point when interoperating with other C code.
And given the lack of bounds checking, you are back at square one.
> I mean Windows API used C.
Windows API used C, because by the time Windows came around UNIX was already well established in the enterprise.
C spread outside UNIX, because many of us were using C dialects, e.g. Small-C, on personal computers when bringing work home from job/university.
And that's exactly why C wins, in my opinion. Because that means interoperability.
Storing a pointer + length pair in a fixed layout struct is bad from a normalization standpoint. They are independent data. If you don't separate them you will end up with data redundancies as soon as you have parallel arrays. If you use dynamic vectors and in C++ or similar languages and you ask yourself which is the object that you should call .size() on, that's when you notice that it's morally wrong to bundle pointer + length.
Or more precisely, 68% of Linux kernel exploits according to Google's talk at Linux Kernel Summit 2018.
You can't rip them out and replace them with something better without massive code legacy issues. And then you still haven't dealt with other sources of buffer overflows in C.