Well, the majority still seems to be unhappy. I think it is mainly because I chose Linux as an example and it reads like I think that Linux is not secure software which is not at all what I intended to say and also (obviously?) not what I think is true.

I agree because it is exactly what my claim is. It would still be foolish to say that open source software is by design more secure than proprietary. I know that this is not what you said and you most likely also don’t mean that, but there are enough people who think that way because they read everywhere that OSS=secure software.

Your example with xz however does not really hold imo. The xz bug was not found because xz is open source but because someone realized, that their ssh session build up took longer than usual and they then used valgrind to check for issues and not because they looked in the source code. It wasn’t even really an easy to spot backdoor because it was a malicious compressed file that changed the build process while running the tests and injecting the actual backdoor in the compiled file. Therfore this would have been found with proprietary software with the same likelyhood.

And regarding my analogy: I also like it more when things are recyclable, that is also why I like open source software more and have more trust in it. But now that I think about it, that wasn’t the best analogy I could’ve chosen but it was the first thing that came to my mind.

Could we please stop associating open source with security? Don’t get me wrong, I love open source software and it is easier to trust open source software than proprietary, because it is highly unlikely, that they hide stuff like trackers in there. It is also most of the time highly configurable and sometimes even hackable and as a software developer you are able to look into the mechanisms behind the APIs which is sometimes really helpful.

But events like the lzma incident last year and predictable openssl RNG in Debian some time ago (https://lists.debian.org/debian-security-announce/2008/msg00152.html) should tell us, that open source doesn’t mean secure software. And the argument, that there are many people looking at the code is not really true. E.g. many maintainers of the linux kernel only look at specific parts/drivers in it and maybe into some other things they need for that. There are probably only a few people if any (apart from governments), that have read, understood and analyzed the linux kernel in its entirety with all the (open source) drivers built into it and all the possible combinations of configurations. And I don’t want to know how many have done all that for less popular projects. And even if that is done at some point for an upstream project, you would have to check the patches from your distro and if there are any do it all for yourself again. And when the next release arrives you would have to do all that in its entirety again (although with some head start) if a new version arrives (that has, say, at least a thousand lines of code changed, removed or added). And now think about how many big releases come with some software per year. And don’t forget to also include all the dependencies you have to check including the compiler and standard library of the language(s) used.

Of course it is easier to do all that for OSS as an outside party because you don’t have to decompile it, but it is still increadibly hard. And only to be easier to analyze for security risks doesn’t mean to be more secure just like packaging being recyclable doesn’t mean that it will be recycled.

For the purpose of algorithm verification, the final and/or pushdown automaton or probably sometimes even Turing Machines are used, because they are easier to work with. “Real” regular expressions are only nice to write a grammar for regular languages which can be easily interpreted by the computer I think. The thing is, that regexs in the *nix and programming language world are also used for searching which is why there are additional special characters to indicate things like: “it has to end with …” and there are shortcuts for when you want that a character or sequence occurs

• at least once,
• once or never or
• a specified number of times back to back.

In “standard” regex, you would only have

• () for grouping,
• * for 0 or any number of occurances (so a* means blank or a or aa or …)
• + as combining two characters/groups with exclusive or (in programming, a+ is mostly the same as aa* so this is a difference)
• and sometimes some way to have a shortcut for (a+b+c+…+z) if you want to allow any lower case character as the next one

So there are only 4 characters which have the same expressive power as the extended syntax with the exception of not being able to indicate, that it should occur at the end or beginning of a string/line (which could even be removed if one would have implemented different functions or options for the tools we now have instead)

So one could say that *nix regex is bloated /s