I think this is the breaking point where replacing our code written in C for code written in memory safe languages is becoming urgent.
The vast majority of vulnerabilities found recently are directly related to being written in memory unsafe languages, it's very difficult to justify that a DNS/DHCP server can't be written in rust or go and without using unsafe (well, maybe a few unsafe calls are still needed, but these will be a very small amount)...
They really are, though. Security is all-encompassing, including not just programming languages, libraries, programs, but also systems, humans and their processes. Don't forget physical security either.
There are no silver bullets, and if the Rust Hype Squad told you there were and all you had to do was buy their product, they were just bamboozling you to push adoption of their pet language.
Write in whichever language you like. Including Rust. Including C. Even PHP. You can write secure software if you put your mind to it.
That's not a great analogy. There are no pick-proof locks (...that are mechanically operated and admit keys; I'm presuming you didn't mean "pick-proof" in that there is no keyhole to pick but is defeated some other way, e.g. a keypad, you meant maximally pick-resistant in contrast to "easily picked")
And honestly? No! If someone can kick your door down, don't waste your money on a super-secure lock, they will just kick your door down. And if you're having your door kicked down on the regular, don't even focus on bolstering the door (they'll either start using power tools or take some other tact like smashing your windows or drilling through the floor or roof). Leave the door open if you like, but move whatever's attracting the attention of intruders somewhere more defensible. Focusing on the securing the wrong thing is also a security flaw!
Ah...these security issues are perfectly cromulent because my shiny new magical beans language let's me screw up in brand new ways, not the old and busted boomer screwups. New is always awesome.
How many CVEs in coreutils over the years? The project has the advantage of being old enough for them to be fixed. Call me when the rust rewrite has been there that long and still has more CVEs than the GNU counterpart.
Maybe coreutils is so old that most security vulnerabilities was solved before CVE even existed. But I think this is also a good argument why we are replacing a solid piece of C code to Rust just because it is "memory safe" and then have lots of CVEs related to things like TOCTOUs (that Rust will not save you).
I'm not against rewriting it in Rust because I believe it really may help in certain class of bugs, but indeed it should not be replacing the old version instantly for that reason. Both could co exist, even tho you still need some guinea pigs to test it out and find issues.
Other than security, Rust brings major improvement to the tooling and may help bring fresh members that wouldn't want to contribute to C code. I understand why some projects go that route
> Other than security, Rust brings major improvement to the tooling and may help bring fresh members that wouldn't want to contribute to C code. I understand why some projects go that route
But it loses old members who don't program in rust, already know the projects, all the reasons of why "this thing" was done "that way". and introduces a new set of bugs, plus now you have two versions of the same thing to maintain.
People thinking that using a superior tool (on paper) enables them to automatically write better tools than the ones who are battle tested over the years baffles me to no end.
Yes, you can go further, possibly faster. OTOH, nothing replaces experience and in-depth knowledge. GNU Coreutils embodies that knowledge and experience. uutils has none, and just tries to distill it with tests against the GNU one.
...and they get 44 CVEs as a result in their first test.
There was an article posted to HN recently that enumerated bugs in the rust rewrite.
Iirc the bugs had to do with linux system details like fs toctou and other things you'd only find out about in production.
Ideally we'd have a better way of navigating platform idiosyncrasies or better system APIs, so that every project doesn't have to relearn them at runtime. But the rewrite isn't pure downside.
My read on those was basically that the classic filesystems are hopelessly broken and we need ACID guarantees in the next-gen filesystems, like 20 years ago.
Not saying all of them were about FS TOCTOU bugs but once I got to these, that was my takeaway.
Obviously just using Rust cannot fix _all_ bugs, and I reject any criticisms towards Rust rewrites that tear down this particular straw man (its goal being to make it impossible to argue against). That's toxic and I get surprised every time people on HN try to argue in that childish way.
But if we can remove all C memory safety foot guns then that by itself is worth a lot already.
Losing decades-old knowledge on how the dysfunctional lower-level systems work would be regrettable and even near-fatal for any such projects. That I'd agree with. But it also raises the question on whether those lower-level systems don't need a very hard long look and -- eventually -- a replacement.
> The project has the advantage of being old enough for them to be fixed.
That's exactly the point though - replacing established projects is inherently risky no matter how many safety buzzwords the replacement can cram into its GitHub tags.
Maybe the problem is the way we think of dynamic memory. “Oh I don’t know what my maximum size for this is going to be, everything has to be dynamic” Is that really true? Is it really the end of the world for programs to declare maximum acceptable sizes for their inputs, and after that error out or use a ring buffer? If sizes were known you could design around that when using them. Your ram bank is finite, why is every layer inside of it then designed to pretend to be infinite? The rust thing strikes me as a massive waste of time and doesn’t solve the fundamental problem of modeling our programs correctly for reality which is finite system resources, and not just memory. c.f. Chrome loading 4 GB models onto people’s machines.
This is exactly how people thought before 1995. Then everyone started "smashing the stack for fun and profit." In the end, you're trading one set of bugs (dynamic memory bugs and hard to reliably exploit) for another (overflow and easy to reliably exploit).
> Is it really the end of the world for programs to declare maximum acceptable sizes for their inputs, and after that error out
It's supremely annoying when you run into arbitrary limits like that as a user. Often it's like a deliberate expiration date for the software as the world moves on to larger files/etc.
The problem is the lack of talent that is willing to work on this, not the language.
AI Security researchers at least do something. If it was so easy to rewrite everything in rust, I don't know why the response to this incidents isn't a rock solid replacement in rust, the next day.
I tell you why that is. Working on these things doesn't give you stars on github.
That is a very pretentious opinion. Dnsmasq is a ubiquitous project, ~14 years old, and has maintainers that are very experienced in c and in the codebase. Telling them to rewrite in a language they are (maybe) unfamiliar with, even with the help of AI, will make these maintainers' experience worthless.
People seem to think that rewriting in rust just magically fixes all issues, but that's not how it works (See recent uutils CVEs). Rewrites tend to have more bugs because the code is new and hasn't been reviewed as much.
I'm pretty sure we are getting close to the point where a few thousand bucks worth of tokens is enough for an agent coding session to reproduce a significant sized (but not linux kernel sized) C codebase in Rust that's 100% security bug for security bug compatible with the original. And _maybe_ "given enough eyeballs, all bugs are shallow" was true or even close top true once. But non of the "new code" ever has a _single_ eyeball cast over it. You know how sometimes you can stare into the code you wrote for weeks, but as soon as somebody else sees it they go "Hmmm, that bit looks odd. Are you sure it's right?" For most vibe coders or agents coders, it's all the same tool that generated the code that's looking for the bugs - it seems reasonable to assume that if a particular LLM generated the buggy code in the first place, it's at least as unlikely to find the bugs as a human who write buggy code?
> I'm pretty sure we are getting close to the point where a few thousand bucks worth of tokens is enough for an agent coding session to reproduce a significant sized (but not linux kernel sized) C codebase in Rust
Given a comprehensive test suite for the original, probably, yes. if the test suite isn't great, you are still going to spend a lot of time/tokens chasing edge cases.
> that's 100% security bug for security bug compatible with the original
You can do this part without AI. c2rust will give you a translation that retains all the security bugs (and all the memory unsafety). The hope is that the AI in the loop will let you convert it to idiomatic rust (and hence avoid the memory unsafely, and in doing so, also resolve some of the security issues).
> If it was so easy to rewrite everything in rust, I don't know why the response to this incidents isn't a rock solid replacement in rust, the next day.
Meaning that AI/Rust enthusiasts are supposed to supply solutions. Of course they won't.
There is comment like this everywhere, if you don't see them it's just that you don't want to see them. There are a little less frequent than 5 years ago but still frequent enough in each c, c++ or rust discussions.
Which part of the comment comes off as fanatic to you?
I'd disagree with that poster that you can write 100% security bug free code just like that.
> There is comment like this everywhere, if you don't see them it's just that you don't want to see them.
Can you discuss productively without attacks? Mine was, and still is, a question of genuine curiosity. And the only "fanatic" thing in your linked comment is a bogus 100% claim. I'm not seeing fanaticism.
The question is whether the current situation is a short burst of action, and once those most critical bugs get fixed the hype around AI vulnerability scanning will die down, or whether the current crop of system/infra software written in vulnerable languages like C are beyond redemption and they will provide an endless source of critical bugs for AI to find until we fix them by rewriting them in Rust/Go/whatever.
You're likely joking, but in case someone else misunderstands; this is not going to work. Rust with unsafe{} is the only thing you can translate directly to, even with LLMs. Rust with extensive unsafe{} is not something anyone wants to debug or maintain, and is near impossible to improve quickly.
It's more of a good thing that, in most cases, it's on devices that won't send it any packets unless a client first authenticates to a Wi-Fi station or physically plugs into an Ethernet port.
> For a number of reasons, I feel that the only way we got here was via some kind of infinite improbability drive.
Oh very much so! In my mind, it seems that someone must have figured out what the universe was for, and now it's been replaced with something even more bizarre and inexplicable.
My own MaraDNS has been extensively audited now that we’re in the age of AI-assisted security audits.
Not one single serious security bug has been found since 2023. [1]
The only bugs auditers have been finding are things like “Deadwood, when fully recursive, will take longer than usual to release resources when getting this unusual packet” [2] or “This side utility included with MaraDNS, which hasn’t been able to be compiled since 2022, has a buffer overflow, but only if one’s $HOME is over 50 characters in length” [3]
I’m actually really pleased just how secure MaraDNS is now that it’s getting real in depth security audits.
Well, as you bundle Lua 5.1 (as Lunacy), instead of making a library and loading it, and you bundled the 2012 version, you're probably affected by CVE-2014-5461 and others. Lua hasn't been security fix free.
Now, I should probably explain why I’m using Lua 5.1 instead of the latest “official” version of Lua. Lua has an interesting history; in particular Lua 5.1 is the most popular version and the version which is most commonly used or forked against. Adobe Illustrator uses Lua 5.1, and Roblox uses a fork of Lua 5.1 called “luau”. LuaJIT is based on Lua 5.1, and other independent implementations of Lua (Moonsharp, etc.) are based on versions mostly compatible with Lua 5.1.
Lua 5.1 has a remarkably good security history, and of course I take responsibility for any security bugs in the Lua 5.1 codebase since I use the code with the relatively new coLunacyDNS server (Lua 5.1 isn’t used with the MaraDNS or Deadwood servers).
Lua 5.1 is used to convert documentation, but those scripts are run offline and the converted documents are part of the MaraDNS Git tree.
Yeah, I've had patches submitted to Moonscript, Fengari, and luau. Don't need to sell on why 5.1 is useful. Each version is a new lang, not just a few fixes or niceties.
I'm not convinced that vendoring, instead of embedding, is the right way.
The patch landing in 2021, instead of 2014, being one of those concerns.
(And you might want to recheck your assumption of how big 'int' will be, for rg32. C defines it in terms of minimum size, not direct size. int16_t isn't necessarily an alias.)
And, yes, this can be easily checked with a tiny C program:
#include <stdint.h>
#include <stdio.h>
int main() {
uint32_t foo = 0xfffffffd;
uint64_t bar = 0xfffffffd;
uint32_t a = 0;
for(a=0;a<20;a++) { printf("%16llx:%16llx\n",foo++,bar++); }
return 0;
}
If there’s a system where uint32_t is 64 bits, that’s a bug with the compiler (which isn’t following the spec), not MaraDNS.
Are you going to make any other negative false implications about MaraDNS? Because you’re making a lot of very negative accusations without bothering to check first.
Edit: Here’s a version of the above C program which works in
tcc 0.9.25:
#include <stdint.h>
#include <stdio.h>
void shownum(uint64_t in) {
int32_t a;
for(a=60;a>=0;a-=4) {
int n = (in >> a) & 0xf;
if(n < 10) {printf("%c",'0'+n);}
else {printf("%c",'a'+(n-10)); }
}
return;
}
int main() {
uint32_t foo = 0xfffffffd;
uint64_t bar = 0xfffffffd;
uint32_t a = 0;
for(a=0;a<20;a++) {
shownum(foo++);
printf(":");
shownum(bar++);
puts(""); }
return 0;
}
Which is the point. In 2020, when you added Lua, you added a vulnerability that had officially been fixed for six years. Because you vendored, and did not depend on any system package.
I don't think you should take responsibility for bugs. Debugging is the act of removing bugs, so programming is the act of putting them in. Free software, per license, must not come with guarantees because the legal implications of any warranty whatsoever could cripple free software projects. The license of your nice DNS server includes the "as-is" wording for good reason.
That said, it is comendable that you've gone out of your way to hold up your software against very high standards. That kind of quality speaks for itself :)
Very good question. I can tell you why I chose Lua 5.1 for MaraDNS:
• Lua 5.1 is smaller than Lua 5.4
• Lua 5.1 is LuaJIT compatible; Lua 5.4/5.5 isn’t as compatible
LuaJIT is a version of Lua 5.1 which is an incredibly fast scripting language because it, in real time, compiles Lua 5.1 code in to native instructions. The only wart LuaJIT has is that its RISC-V port is incomplete, but that will undoubtedly change as RISC-V slowly gets more popular.
The other reason to stick to Lua5.1 is because Lua changes its syntax between versions; e.g. bitwise operations in Lua 5.4 are very different than how they are done in Lua5.1, to the point it’s difficult to make a polyglot library which can do bitwise operations in both Lua 5.1 and Lua 5.4. I am of the opinion Lua 5.3 should had been named Lua 6.0 for the simple reason that having native integers in Lua is a pretty significant backwards compatibility breaking change.
Since Lua (well, Lunacy) is the only tool in MaraDNS which isn’t standardized (e.g. MaraDNS uses only POSIX-comatible shell scripts, it uses “make” because that’s a standardized tool with multiple implementations, C is also a standard with multiple implementations, etc.), sticking to Lua5.1 allows me to use a version of Lua with multiple implementations and, as such, is informally standardized.
Unless the service accepts Lua code from the internet (and that would be a completely insane thing), the CVE-2014-5461 will not apply. And while I have not reviewed every Lua CVE, I bet most (all?) of then require a specifically crafted code, or at least highly-complex user input (such as arbitrary json)
It's important to look at the actual vulnerability at the context, and not just list any CVE which matches by version.
I should explain how MaraDNS uses Lua 5.1 (actually, Lunacy, my own fork with security bugs fixed as well as security hardening—including, yes, a patch against CVE-2014-5461), so you can get an idea of its attack surface.
MaraDNS has three components:
• MaraDNS, the authoritative server, which goes back all the way to 2001
• Deadwood, the recursive server, which was started back in 2007
• coLunacyDNS, which allows a DNS server to use Lua scripting; this didn’t exist until the COVID pandemic
Neither MaraDNS nor Deadwood use Lunacy (except as a scripting engine for converting documents); only coLunacyDNS uses Lunacy. coLunacyDNS uses a sandboxed and security hardened version of Lunacy (and, yes, I would accept bugs where someone could escape that sandbox), and the Lua scripts which coLunacyDNS uses can only be controlled by a local user and there is no capability to run Lua scripts remotely.
You weren’t replying to me. The parent poster made a good point—a vulnerability in Lua doesn’t mean software running Lua can necessarily be exploited—but, more to the point, I do update Lunacy and make sure it’s secure, just as I still take responsibility for verified important security holes in MaraDNS.
> It's important to look at the actual vulnerability at the context, and not just list any CVE which matches by version.
Unfortunately, that's not enough. Even if the vulnerable parts of the code are not being built, heck even if they have been completely erased from the source code, the auditors will still insist that you're vulnerable and must immediately upgrade, or else they will give your software a failing grade.
I have several libraries that I've written. Not one single serious security bug in them has been found since 1991. Granted, nobody uses my libraries...
Not to diminish your team's achievement! :D But it's important to contextualize claims like this with information about what your userbase looks like
A lot of security and other audits have been performed against it though; MaraDNS, after all, is notable enough to have a Wikipedia page and hundreds of GitHUB stars.
For example, when the Ghost Domain Name DNS vulnerability was discussed, MaraDNS was audited and named (MaraDNS was immune to the security bug, for the record)
Way fewer people are going to look at obscure things, so a lower percentage of issues will likely have been found. There is less fame and fotune in spending security research time on obscure software. Most small libraries won't be covered by any bug bounty programs either for example.
Even accepting the premise, is it not immediately obvious to you that folks will be spending more money and effort aiming AI at higher-impact targets? This isn’t all-or-nothing.
I remember being delighted finding maradns as an alternative to the “do everything” of dnsmasq way back when I set up a dns server, and more importantly, I haven't had to think about it since then.
Agreed, it made a lot more sense to write MaraDNS in C in 2001 though.
The main advantage of writing in C over Rust here in 2026 is that C has two different Lua interpreters, and there isn’t a port of Lua to Rust yet; [1] yes, there are ways to use the C version of Lua in Rust, but that’s different.
If I were to write a new server today, I could very well write it in Go, then use GopherLua for the Lua engine:
Flagged because this discussion about dnsmasq and another dns resolver implementation that has relatively no rollout worldwide by comparison is pointless.
dnsmasq has served me well for like an eternity in multiple setups for different use cases. As all software it has bugs. And once located those get fixed. Its author is also easy to communicate with.
Why should I switch over to something way less proven? I'm quite sure your software also has bugs, many still not located. Maybe because it's less popular/ less well known nobody cares to hunt for those bugs? Which means even if the numbers of found bugs is less in your software at the moment, and it may look more audited for this reason, it may actually be way less secure.
"All software has bugs" is the most meaningless statement ever. It is just used for bonding with fellow bug writers who sit at a virtual campfire and muse about inevitabilities.
Demonstrably some software has fewer bugs, and its authors are often hated, especially if they are a lone author like Bernstein. Because it must not happen!
Projects with useless churn and many bug reports are more popular because only activity matters, not quality.
If DJB is "hated", it isn't because he's a lone author (Linus Torvalds was once a lone author and I don't think he was hated). It's because he can be an asshole. To quote George Bernard Shaw, “The reasonable man adapts himself to the world: the unreasonable one persists in trying to adapt the world to himself. Therefore all progress depends on the unreasonable man.”
DJB is a lot of things, and I have great respect for him, even though I feel he didn’t responsibly maintain Qmail/DJBdns/Publicfile. He made MaraDNS more secure because I carefully read his documentation—I got the idea to have a random source port to give MaraDNS more security from him, which means MaraDNS was unscathed when DNS spoofing was independently discovered in 2007.
The point DJB made was this: It was possible for a skilled C programmer to make a server with few security holes. Even though that’s not as relevant now, with Rust having most of the speed of C and security built in, it did make the Internet a safer place for many years. I remember using Qmail and DJBdns to make the servers at the small company I worked for at the time more secure.
"All software has bugs" so "be wary of the one trying to say they haven't had any in 3 years" not so "I guess all are equal". For extremely low security bug rates either the scope is extremely narrow, the claim is dubious, or the project is a massive effort which the community talks about directly in posts rather than plugs (e.g. curl).
DJB, with Qmail and DjbDNS (as well as Publicfile, which didn’t catch on in an era of CGI scripts), showed that one could have (mostly) security bug free software without the scope being “extremely narrow”, and without the claim being “dubious”.
It’s not normal for software to be so poorly written, one doubts the claim that a security bug hasn’t been found in over three years. If one thinks the claim of no security bugs of consequence in three years is dubious, feel free to do a security audit of MaraDNS (or DjbDNS, which I also will take responsibility for even though my software is, if you will, a “competitor” to DjbDNS), and report any bugs you find.
Speaking of DJB, DjbDNS has had a few security bugs over the years (but not that many), but I’m maintaining a fork of DjbDNS with all of the security bugs I know about fixed:
I am saying all this as someone who has had significant enough issues with DJB’s software, I ended up writing my own DNS server so I didn’t have to use his server (I might not had done so if DjbDNS was public domain in 2001, but oh well).
(As a matter of etiquette, it’s a little rude to claim someone is saying something “dubious”, especially when the claim is backed up with solid evidence [multiple audits didn’t find anything of significance in the last year, as I documented above], unless you have solid evidence the claim is dubious, e.g. a significant security hole more recent than three years old)
People here don't know that MaraDNS was already popular on extremely critical security mailing lists that basically hated anything but qmail and postfix. If you introduce more bugs and blog about them, it will probably gain in popularity. :)
> It’s not normal for software to be so poorly written, one doubts the claim that a security bug hasn’t been found in over three years.
Can you back that claim up with at least some sort of theory? Because it doesn't match my perception of the real world, nor does it match my mental model of how CVEs happen.
I never used Qmail, so I won't comment on it, but I will say I absolutely consider djbdns narrow in scope as well (before accounting the Unix approach, utilized perhaps even more than in MaraDNS, to break that already narrowed scope down into even more focused binaries).
I had believed (and continue to hold) DNS software containing, e.g., an authoritative DNS server which lacks native TCP or DNSSEC support falls squarely into the "narrowly scoped" bucket and would appreciate if you'd not try to decide my opinion for me on any given project in the future.
When you offer up an alternative as technically superior in some manner then yes, it is on you to demonstrate such a claim in a convincing manner. "No bugs in 3 years in this software with a much smaller audience and also look AI audits!" comes across as off topic shameless self promotion. At least if an insightful technical discussion ensued the subthread might prove worthwhile but so far it's just the usual tired shit flinging.
I have far more evidence of a very good security record with MaraDNS than “No bugs in 3 years in this software with a much smaller audience and also look AI audits!”
• The software has been around for 25 years
• The software is popular enough to have been subjected to dozens of security code audits, including two audits in the post-AI era
• In those 25 years, only two remote “packet of death” bugs have been found
• Also, in those same 25 years, only one single bug report of remotely exploitable memory leaks has been found
This isn’t something which, as implied here, has a lot of security bugs only because no one has used or audited the software. This is a long term, mature code base which has only had a few serious security bugs in that timeframe.
For what it's worth I didn't know about maradns prior to this. Maybe it actually sees fairly wide use? Whether or not I accept your evidence would hinge on that. Regardless I think my point stands - if you don't lead with a convincing line of reasoning all that's left is an empty assertion. Unless I happen to recognize you as an authority in the field that's not going to do anything for me since by default you're some stranger on the internet that might be a dog for all I know.
To illustrate the issue with an extreme example, consider that a disused repository on github full of security holes is highly unlikely to have any CVEs regardless of age. The software has to present a worthwhile target (ie have a substantial long term userbase) before anyone will bother to look for exploits. (I guess that might change in the near future thanks to AI but I don't think we're there just yet.)
> dnsmasq has served me well for like an eternity in multiple setups for different use cases. As all software it has bugs. And once located those get fixed. Its author is also easy to communicate with.
I concur. The last part, however, is quite worrisome. Dnsmasq is ran by one person, published on their own git and I did not see any information about other maintainers.
It is a super important (and great, and useful, and everything) software and i have fears of what will happen one day.
Sure, someone can clone and push to github but it may seriously fragment the ecosystem.
They can block traffic to update servers so the computers behind the router aren't all patched up, then exploit them. They also get access to all the IoT devices on the internal network. They can also use your router as a proxy so their scraping/attack traffic comes from your IP address instead of theirs.
If you blindly TOFU ssh sessions, those can be pwned easily in many common use cases. Legacy software configurations like NFS with IP authentication will be bypassed. Realistically the most likely scenario is using your home as a VPN, or a DDOS node.
they could try and exploit any device on your network, and since they see which servers you connect to and how often you communicate with one they can write phishing mails which are tailored just for you.
The AI bug report tsunami is not in all projects. As the top comment notes, MaraDNS didn't have any. I assume djbdns and tinydns didn't either, otherwise they'd shout it from the rooftops.
I never understood why some projects get extremely popular and others don't. I also suspect by now that the reports by tools that are "too dangerous to release" scan all projects but selectively only contact those with issues, so that they never have to admit that their tool didn't find anything.
Maybe this is the kick in the ass Debian needs to upgrade the embarrassingly ancient dnsmasq in "stable" because while I can't think of any new features, the latest versions contain many non-CVE bug fixes.
But I doubt it, they will lazily backport these patches to create some frankenstein one-off version and be done with it.
Before anyone says "tHaT's wHaT sTaBlE iS fOr": they have literally shipped straight-up broken packages before, because fixing it would somehow make it not "stable". They would rather ship useless, broken code than something too new. It's crazy.
They're not going to put a newer version in stable. The way stable gets newer versions of things is that you get the newer version into testing and then every two years testing becomes stable and stable becomes oldstable, at which point the newer version from testing becomes the version in stable.
The thing to complain about is if the version in testing is ancient.
Yeah was about to comment, parent says "if it is ancient", it is not. So the root comment is nothing burger. Stable has 1 release cycle old, and depending on how things play out, testing may have 2.93 or later anyways.
That whole model dates to before automated testing was even really a thing, and no one knew how to do QA; your QA was all the people willing to run your code and report bugs, and that took time. Not to mention, you think the C of today is bad? Have you looked at old C?
And the disadvantage is that backporting is manual, resource intensive, and prone to error - and the projects that are the most heavily invested in that model are also the projects that are investing the least in writing tests and automated test infrastructure - because engineering time is a finite resource.
On top of that, the backport model heavily discourages the kinds of refactorings and architectural cleanups that would address bugs systemically and encourage a whack-a-mole approach - because in the backport model, people want fixes they can backport. And then things just get worse and worse.
We'd all be a lot better off if certain projects took some of the enthusiasm with which they throw outrageous engineering time at backports, and spent at least some of that on automated testing and converting to Rust.
> That whole model dates to before automated testing was even really a thing, and no one knew how to do QA; your QA was all the people willing to run your code and report bugs, and that took time.
That's not what it's about.
What it's about is, newer versions change things. A newer version of OpenSSH disables GSSAPI by default when an older version had it enabled. You don't want that as an automatic update because it will break in production for anyone who is actually using it. So instead the change goes into the testing release and the user discovers that in their test environment before rolling out the new release into production.
> On top of that, the backport model heavily discourages the kinds of refactorings and architectural cleanups that would address bugs systemically and encourage a whack-a-mole approach - because in the backport model, people want fixes they can backport.
They're not alternatives to each other. The stable release gets the backported patch, the next release gets the refactor.
But that's also why you want the stable release. The refactor is a larger change, so if it breaks something you want to find it in test rather than production.
You're going to have to update production at some point, and delaying it to once every 2 years is just deferred maintenance. And you know what they say about that...
So when you do update and get that GSSAPI change, it comes with two years worth of other updates - and tracking that down mixed in with everything else is going to be all kinds of fun.
And if you're two years out of the loop and it turns out upstream broke something fundamental, and you're just now finding out about it while they've moved on and maybe continued with a redesign, that's also going to be a fun conversation.
So if the backport model is expensive and error prone, and it exists to support something that maybe wasn't such a good idea in the first place... well, you may want something, but that doesn't make it smart.
> You're going to have to update production at some point, and delaying it to once every 2 years is just deferred maintenance. And you know what they say about that...
Updated what, specifically in production?
If you need a newer version of Python or Postgres or whatever it is possible to install it from third-party repos or compile from source yourself. But having a team of folks watch all the other code out there is a load off my plate: not worrying about libc, or OpenSSH, or OpenSSL, or zlib, or a thousand other dependencies. If I need the latest version for a particular service I would install that separately, but otherwise the whole point of a 'packagized' system is to let other folks worry about those things.
> So when you do update and get that GSSAPI change, it comes with two years worth of other updates - and tracking that down mixed in with everything else is going to be all kinds of fun.
I've done in-place upgrades of Debian from version 5 to 11 at my last job on many machines, never once re-installing from scratch, and they've all gone fine.
Further, when updates come down from the Debian repos I don't worry about applying them because I know there's not going to be weird changes in behaviour: I'm more confident in deploying things like security updates because the new .deb files have very focused changes.
One is security updates and bug fixes. These need to fix the problem with the smallest change to minimize the amount of possible breakage, because the code is already vulnerable/broken in production and needs to be updated right now. These are the updates stable gets.
The other is changes and additions. They're both more likely to break things and less important to move into production the same day they become public.
You don't have to wait until testing is released as stable to run it in your test environment. You can find out about the changes the next release will have immediately, in the test environment, and thereby have plenty of time to address any issues before those changes move into production.
If you don't like the debian model, didn't use debian. There are people that like the debian model, it seems like you aren't one of them, though. That doesn't make them wrong.
> You're going to have to update production at some point, and delaying it to once every 2 years is just deferred maintenance. And you know what they say about that...
Doing terrible work every 2 years is better than doing it every day?
Get thru the issues once every 2 years is entirely fine. Farther than that and you get problems. We do that for ~500 systems of very varied use. I wouldn't want to do it yearly (or dread on rolling release) but I also wouldn't want to do it any less often coz of issues you mentioned.
> And if you're two years out of the loop and it turns out upstream broke something fundamental, and you're just now finding out about it while they've moved on and maybe continued with a redesign, that's also going to be a fun conversation.
Having that sprung on you because you decided to run everything on latest is worse.
"Oh we have CVE, we now need to uproot everything because new version that fixes it also changed shit"
With release every year or two you can *plan* for it. You are not forced into it as with "rolling" releases because with rolling you NEED to take in new features together with bugfixes, but with Debian-like release cycle you can do it system by system when new version comes up and the "old" one still gets security fixes so you're not instantly screwed.
> So if the backport model is expensive and error prone, and it exists to support something that maybe wasn't such a good idea in the first place... well, you may want something, but that doesn't make it smart.
It exists in that format because people are running businesses bigger than "a man with a webpage deployed off master every few days"
Clearly you disagree with the debian stable perspective. That's fine, it's not for everyone. You can just run debian unstable or debian testing, depending on where exactly you draw the line.
If you want the rolling release like distro, just run debian unstable. That's what you get. It's on par with all the other constantly updated distros out there. Or just run one of those.
Also, Debian stable has a lifetime a lot longer than 2 years, see https://www.debian.org/releases/. Some of us need distros like stable, because we are in giant orgs that are overworked and have long release cycles. Our users want stuff to "just work" and stable promises if X worked at release, it will keep working until we stop support. You don't add new features to a stable release.
From a personal perspective: Debian Stable is for your grandparents or young children. You install Stable, turn on auto-update and every 5-ish years you spend a day upgrading them to the next stable release. Then you spend a week or two helping them through all the new changes and then you have minimal support calls from them for 5-ish years. If you handed them a rolling release or Debian unstable, you'd have constant support calls.
> You don't want that as an automatic update because it will break in production for anyone who is actually using it
The problem with this take is that it’s stuck in the early 2000’s, where all servers are pets to be cared for and lovingly updated in place.
It’s also circular: you have the same problem with the current model if you don’t have a test environment. And if you do have a test environment, releases can be tested and validated at a much higher cadence.
If you want that, you don't want Debian. Other people do.
Some people will even run Debian on the desktop. I would never, but some people get real upset when anything changes.
Debian does regularly bring newer versions of software: they release about every two years. If you want the latest and greatest Debian experience, upgrade Debian on week one.
From your description, you seem to want Arch but made by Debian?
> From your description, you seem to want Arch but made by Debian?
Isn't that essentially Debian unstable (with potentially experimental enabled)? I've been running Debian unstable on my desktops for something like 20 years.
Refactoring and rewrites prove time and time again that they also introduce new bugs and changes in behaviour that users of stable releases do not want.
For what you want, there are other distributions for that. Debian also has stable-backports that does what you want.
No need to rage on distributions that also provide exactly what their users want.
> That whole model dates to before automated testing was even really a thing, and no one knew how to do QA; your QA was all the people willing to run your code and report bugs, and that took time. Not to mention, you think the C of today is bad? Have you looked at old C
The automatically tested Debian release is called Debian Testing. And it is stable enough.
Debian Stable is basically "we target particular release with our dependencies instead of requiring customer to update entire system together with our software". That model works just fine as long as you don't go too far back.
> On top of that, the backport model heavily discourages the kinds of refactorings and architectural cleanups that would address bugs systemically and encourage a whack-a-mole approach - because in the backport model, people want fixes they can backport. And then things just get worse and worse.
Narrator: It turned out things were not getting worse, they were just fine.
> We'd all be a lot better off if certain projects took some of the enthusiasm with which they throw outrageous engineering time at backports, and spent at least some of that on automated testing and converting to Rust.
That project is RedHat, not Debian, they backport entire features back to old versions (together with bugs!)
Don't get me wrong, I use and encourage extensive automated testing. However only extensive manual testing by people looking for things that are "weird" can really find all bugs. (though it remains to be seen what AI can do - I'm not holding my breath)
You don't have to use Debian stable, if you'd prefer Ubuntu every 6 months, or Fedora (6 months? 9 months?), or even Arch Linux updated daily ...
I use Arch on my laptop, when I got it 2 years ago the amd gpu was a bit new so it was prudent to get the latest kernel, mesa, everything. Since I use it daily it's not bad to update weekly and keep on top of occasional config migrations.
I use Debian stable on my home server, it's been in-place upgraded 4-ish times over 10 years. I can install weekly updates without worrying about config updates and such. I set up most stuff I wanted many years ago, and haven't really wanted new features since, though I have installed tailscale and jellyfin from their separate debian package repos so they are very current. It does the same jobs I wanted it to do 8 years ago, with super low maintenance.
But if you don't want Debian stable, that's fine. Just let others enjoy it.
About a decade ago I switched to Ubuntu LTS because of Debian’s “policy?” of having pretty old packages in “stable” and a long release cycles.
Nowadays, even with Ubuntu’s two year or so release cycle I have to use 3rd party packages to have up to date software (PHP being one) and not some version from three years ago.
We no longer live in a world (with few exceptions) where running a 3-5 year old distribution (still supported) makes sense.
I am running debian oldstable on two rpi-based appliances i built at home. They have been working fine for several years.
I'll have to update them because eventually security updates will stop. That means that the python code on them no longer works on current python versions, C++ needs some tweaks because some library changed API.
Better to do these things every few years than every 6 months for no reason whatsoever.
> That means that the python code on them no longer works on current python versions, C++ needs some tweaks because some library changed API.
And this is why you update often, to keep up with the programming language ecosystem too. I have seen way too many times software unmaintained for years and then when it was actually time to upgrade it would take much more time to bring it to current framework versions than it would have taken if it was updated regularly throughout the years.
And I was not referring to hobby projects you do at home.
It depends on how you look at it. I use Debian stable in the smallet possible configuration because it is, well, stable. A rock on which I put docker to run actually useful services, which are upaded the way I want.
If I was to run dnsmasq on Debian, it would be in a container. Since I run Pihole (in a container), it kinda is.
What if the new release which contains the fixes has new dependencies and those also have new dependencies? I assume they have to Frankenstein packages sometimes to maintain the borders of the target app while still having major vulns patched right in stable.
And no "good faith" assumption here, since you literally claim debian stable ships broken kernels, according to you nobody should be able to even boot a computer.
That's what stable is for though. Like, sure, stable's policy is ludicrous and you would have to be insane to run stable. But the remedy for that isn't to try to change Debian policy, it's to get people to stop running stable. Maybe once no-one uses it Debian will see sense.
Never liked using dnsmasq. Always felt like too much in one tool. A local caching resolver, dhcp server, and tftp/pxe boot setup were always things I preferred to configure separately.
That's kinda the point. It is "i run a small router" app in a box.
DHCP and DNS are connected, PXE requires DHCP entries, so to do a simple setup you'd need to glue together at least 3 daemons otherwise, all with different config syntax
There are few dnsmasq (only?) features that are indispensable to some. Examples: sending query of *.example.com to certain upstream servers, or returning NXDOMAIN for phishing sites, or adding all resolved IPs for *.example.org to an ipset for policy routing. The last one works on FreeBSD as well although BSD does not have ipset. The list of *.example_xyz.com can be huge and it is said recent dnsmasq can handle them efficiently.
I agree, it also goes against the Linux "way of doing things".
For example, Opnsense uses the dhcp portions of dnsmasq only (and unbound for the dns parts) which just feels 'wrong'.
When I first came across Linux you would download the code (very slowly) to /usr/src/linux (extract and cd) and run "make config". You'd answer quite a lot of y/n and later y/n/m questions and then copy a binary and later on run a script to put things in place. Then you would fix up lilo and off you trot ... or not 8)
Is that the Linux way you are on about? No obviously not 8)
I think you mean the "unix idealized but never really happened exactly but we are quite close if you squint a bit ... way" where each tool does one job well and the pipeline takes up the slack.
What is the nature of these findings? There’s a big difference between AI finding a buffer overflow vs. identifying a fundamental protocol flaw. Could AI realistically discover something like the Kaminsky attack? or even something which is an amplification exploit like the NXNSAttack?
For folks with more experience in this specific domain, dumb question: why is more software in this space not written in e.g. Erlang or some other garbage collected, concurrent language runtime?
The initial release of dnsmasq was in 2001. The list of viable languages for a high-performance network server at the time was still not all that long. Erlang wasn't on it. Too big a performance hit, too much opaque runtime that may not have been stable at the time, too few contributors, big dependency footprint of stuff most things wouldn't have installed. (When I used Erlang for a production system in more like the 2015 time frame it still had rough corners if you weren't using it exactly for the use case it was meant for.) This isn't specially a criticism of Erlang, it would have been like this across many languages and runtimes.
A lot of these systems that are getting hit, and will probably continue to be hit over the next few weeks or months, have a similar story. The Linux kernel's only other potentially viable choice was C++ at the time. OpenSSL, a perennial security offender, was started in 1998. You can look up your own favorite major system library with major security issues and it's probably the same story.
I'm as aggressive as anyone about saying "don't write a new project in C for network access", but cast me back to 1998 and I couldn't tell you what other viable choices there are either. There are safer languages, but they were much, much smaller than the C community, and I couldn't promise you how stable they were either. Java was out, and I don't know when to draw the exact line as to when it became a serious contender for a network server, but late 200Xs would be my guess; certainly what I saw in 1999 wasn't yet.
Example: I ran a Haskell network server in 2011 for something relatively unimportant and it fell over under conditions that would not have been very extreme for a production network; I know it was Haskell and not my code because I reused the same code base in 2013 with no changes in the core run loop and it did about 90% better; still not enough that I would have put that system into a real production use case but enough to show it wasn't my code failing. So while Haskell may have existed in the 200Xs, it wouldn't have qualified as a viable choice for a network server at the time.
There's a lot more viable choices today than there used to be.
I was a language explorer and I think I hadn't heard of Ocaml by then. Even if you built the best dnsmasq project ever in Ocaml nobody would have heard of it or known how to build it.
I never liked dnsmasq or the Pi-Hole dderivation and do not use it but many people seem to love this software. I don't think there is any amount of CVEs that could convince people to stop using it
Sorry, I got carried away, the whole blind trust in AI tools gets me wound up a bit too fast, I'll try to be better.
> We've already asked you once.
there are no notifications of any kind about that or the fact the comment has been flagged so I genuinely didn't notice previous one and only noticed this after I noticed rate limiting.
LLMs certainly make it more feasible to rewrite a product in a memory-safe language, eliminating a whole class of bugs.
Flawless software is hard for an LLM to write, because all the programs they have been trained on are flawed as well.
As a fun exercise, you could give a coding agent a hunk of non-trivial software (such as the Linux kernel, or postgresql, or whatever), and tell it over and over again: find a flaw in this, fix it. I'm pretty sure it won't ever tell you "now it's perfect" (and do this reproducibly).
Have you ever met a security engineer? I’ve never met one who was also a good engineer (not saying they don’t exist, I just haven’t met one). Do they find vulnerabilities? Sure. Could they write the tools they use to find vulnerabilities, most probably not.
The CVEs here have their fair share of silly C problems, but also more rigid input validation and handling. These more rigid validations exclude stuff which may even be valid by the spec, but entirely problematic in practice.
As examples, take a look how many valid XML documents are practically considered unsafe and not parsed, for example due to recursive entity expansion. This renders the parsers not flawless and in fact not in spec.
Or, my favorite bait - there should be a maximum length limit on passwords. Why would you ever need a kilobyte sized password?
Just because something is good at finding bugs, it may not find all the bugs. Finding a bug only tells you there was one bug you found, it doesn't tell if the rest is solid.
The vast majority of vulnerabilities found recently are directly related to being written in memory unsafe languages, it's very difficult to justify that a DNS/DHCP server can't be written in rust or go and without using unsafe (well, maybe a few unsafe calls are still needed, but these will be a very small amount)...