WolfSSL sucks too, so now what? | Hacker News Reader

WolfSSL sucks too, so now what?

blog.feld.me

148 pointsthomasjb129 comments

This is the WolfSSL maintainer's response[1]

> This ticket is rather long and has a lot of irrelevant content regarding this new topic. If I need to bring in a colleague I do not want them to have to wade through all the irrelevant context. If you would like, please open a new issue with regards to how we support middlebox compatibility.

The author turns this into:

> The GitHub issue comment left at the end leads me to believe that they aren't really interested in RFC compliance. There isn't a middleground here or a "different way" of implementing middlebox compatibility. It's either RFC compliant or not. And they're not.

This is a bad-faith interpretation of the maintainer's response. They only asked to open a new, more specific issue report. The maintainer always answered within minutes, which I find quite impressive (even after the author ghosted for months). The author consumed the maintainer's time and shouldn't get the blame for the author's problems.

[1]: https://github.com/wolfSSL/wolfssl/issues/9156

I don't know, I don't think it's really a huge waste of time considering I just read the entire comment thread in a handful of minutes. And beyond that, failing to comply with RFC requirements is the bug here -- a workaround existing for a specific language isn't a fix.

Again: the maintainer does not say there is no bug. He says: please open a new issue, with a proper title and description for the actual underlying problem. Is that seriously too much to ask? Instead, the guy writes a whole blog post shitting on the project. Does anyone still wonder why people burn out on maintaining FOSS projects?

Not great behavior I agree, but what else is there to say other than "it does not match the spec at point 1.2.3"?

Then opening the ticket should be easy enough?

I certainly understand the maintainer here, because that’s what I keep telling colleagues at work.

Tickets get really cumbersome if they are not clear and actionable.

> Then opening the ticket should be easy enough?

For both of them! Since both of them are aware now, either one could open that ticket. If the maintainer has very specific ideas about how a ticket should look, maybe they can do that themselves quickly, now that they are aware of not complying with the RFC. Then the ticket will perfectly match their expectations.

Because that's incredibly entitled. The maintainer is already the one who has to fix it.

The maintainer is usually also the one who has to trace the root cause, which in this case the issue reporter did, which is certainly more work than creating an issue according to the formatting and other requirements the maintainer may have. So in that light, the reporter of the issue already did a big chunk of work for the maintainer or the project. I wouldn't really call them acting "entitled" after that. Clearly they put in effort more than could be expected already.

...that's what they are asking, yes.

Exactly, that's all his PR had to be. The history of finding the issue could be an interesting story (I bet it involves Elixir!), but in places it reads as almost malicious. If I received a PR anything like that on something I maintained, it would be received very poorly. The author comes off as overly aggressive toward the maintainers and far too sensitive to their response.

Open Source is not Free Support, the sooner this reality sets in (accelerated surely by AI spam) the sooner we get to the happy place.

It's pretty standard to open a new issue and reference the previous issue for context, while keeping the new issue specific about what needs to be addressed - ie. RFC compliance.

I don't see the problem here at all - it was a reasonable request and it would have taken `feld` all of 2 minutes to do. Certainly less time than writing that blog post.

It's not entirely WolfSSL's fault. TLS 1.3 is a mass of kludges and hacks to deal with the fact that they created a new protocol that's nothing like TLS 1.0-1.2 but dressed it up to make it look like TLS 1.2. It even lies about its protocol version in the handshake, hiding the real version in one of the many extensions they had to invent to kludge it into working. And in terms of RFC compliance, one of the most widely-used implementations isn't compliant, it doesn't send any of the mandatory-to-implement cipher suites in its client hello which means unless you want to trigger a rehandshake on every single connect you have to implement their non-compliant form of TLS 1.3.

The real problem though is that they made a protocol that really, really wants to pretend it's TLS 1.2 when it really isn't anything like TLS 1.2. I wouldn't blame "middleboxes" for getting confused when they encounter that.

> wants to pretend it's TLS 1.2 when it really isn't anything like TLS 1.2.

I've seen a ton of this recently as Amazon has the option for TLS 1.3 with post quantum encryption on cloudfront now. A whole ton of different middleware shits itself.

The problem is there are many middleboxes that monitor port 443 and will drop any traffic that they can't decode as TLS (which in this case means TLS 1.2 or below). The choice was between masking traffic as an earlier version of TLS or forcing the replacement of all of those middleboxes. It's a no-brainer.

Then don't put it on 443 and pretend (badly) that it's TLS 1.2. Given that QUIC also uses 443 (and 80) without too many problems and that doesn't look anything remotely like TLS, presumably non-TLS 1.2 traffic to 443 is OK.

The problem isn't really the port used, it's the uncanny-valley approach they took in creating something that looks like a creepy zombie version of TLS 1.2, which keep-suspicious-things-out appliances quite rightly get suspicious over.

But QUIC doesn’t use 443/TCP; it uses 443/UDP. So it’s unsurprising that middleboxes that care about 443/TCP would ignore it. That doesn’t support your claim that “non-TLS 1.2 traffic to 443 is OK.”

The point I was trying to make, probably badly, was that there was no need to make TLS 1.3 pretend to be TLS 1.2 going to TCP/443. They could have picked some new port, called it TLS 2.0 (which is what it actually is), and run with that. If QUIC can pick its own port and, by the looks of it, not run into massive problems, there's no reason why TLS 2.0 can't do so too.

The maintainer should just open a new issue for RFC compliance himself since that's a pretty big issue and he obviously thinks OP spams too much.

This game of stalling / obfuscating via the issue tracker gets very old.

> The maintainer should just

Out of interest: which FOSS projects are you maintaining, and how many users do these have, approximately?

Out of interest, how is that relevant? Are we not able to criticize a FOSS maintainers response unless we run a project of scale ourselves? The maintainer is clearly engaging and knows what the problem is but stalls on the "last mile" which is issue creation. Do you agree?

wolfSSL also sells commercial licenses so it's not like they're going uncompensated for their work. Regardless, we shouldn't put people on pedestals because their title is "FOSS maintainer"

Unless you're paying you are not entitled to anything apart from forking and fixing it yourself.

You are especially not entitled to bullying maintainers as has been unfortunately the standard in infosec.

Open source is not about you.

https://gist.github.com/richhickey/1563cddea1002958f96e7ba95...

IMO more projects have to explicitly state this for example in a terms document, like https://github.com/mhoye/maintenance-terms/blob/main/MAINTEN...

> Open source is not about you.

You know a social movement went full circle when a criticism that is so scathing, you couldn't have possibly come up with it and make it trend before, even if you gave it your all, is now a motto and a point of pride for those who follow it.

This is happening at the same time where hundreds of millions of regular variety consumers are being fed propaganda daily about how it's "finally time to switch to Linux", because it's so much better for them, the individual. If only they knew it's apparently not actually about them, never has been, and never will be.

When exactly is 'before'? Before Github existed to put front and center your code and its issues? Before it became an expectation to have a a rich Github profile when you're considered for a job position?

Of course I wouldn't have been able to come up with this statement because the perverted view of OSS devs owing free work to the users of their software was not so pervaisive.

On your edit: a bit rich saying the calls for switching to Linux propaganda, especially with the downturn of UX of windows and macos... Also why just hundreds of millions.. Go for hundreds of billions if you're just going to pull out numbers. Apart from that - even if Linux is not about the users, it is in many cases better for them as-is. Funny how that works with no conflict.

> When exactly is 'before'?

"Exactly"? I'm afraid that's not a very physically sound request. But let's say, prior to 2026-02-14T03:46:03Z then. I hope that suffices.

> Of course I wouldn't have been able to come up with this statement

That would make sense, because you specifically I never expected to: https://en.wikipedia.org/wiki/Generic_you

> Also why just hundreds of millions.. Go for hundreds of billions if you're just going to pull out numbers

You see, that would be because I did not just pull out an arbitrary number. "How many Windows users there are" is a reported fact you can just search for, and even the total is not "billions" (plural). I know, I was surprised too. From the horse's mouth: https://blogs.windows.com/windowsexperience/2025/06/24/stay-...

My first comment on this site pointing out that a FOSS user sounds entitled is from 2021. I've been saying it outside the site for 10+ years, spanning back to the time when it wasnt cringe to have a Github sticker on your laptop.

I maintain several FOSS projects, although none as popular as wolfssl and if I want to make a new issue to make it more clean, I usually do it myself, because then I can write it the way I want, and include the information, and only the information, that I think is important. If I ask someone else to do it, there's a pretty good chance they won't write it the way I would like, if they write it up at all.

That's actually impossible to answer. I maintain or contribute to or have contributed to several FOSS projects whose number varies depending on how you want to count them, and neither myself nor anyone else who contributes to any FOSS project has the faintest idea how many people use them, especially if they're included in widely-used distros where the number is anything from zero to $number_of_distro_users.

Why should that be the maintainer's burden?

If they're doing this and bothering to interact with tickets at all, presumably they've willingly taken on a duty to the software's quality and all that that entails.

Maintaining an open-source software project is frequently a hobby that’s performed out of a labor of love. There’s no duty owed to anyone, nor should one be implied by past behavior. The open-source community is not a slave trade.

Hence why I said "voluntarily".

I don't see how that changes anything (and you didn't say "voluntarily"). Volunteering does not create a duty. One can volunteer to pick up litter and give up halfway through; the only consequence would be disappointment.

> Volunteering does not create a duty.

Volunteering to maintain a project literally does create entail accepting duties, that's what taking on the role of maintainer entails. They are of course free to give up that role at any time, but those duties exist while that role has been adopted.

Let's suppose for the sake of the argument that while you are a volunteer, you take on some duty. What is the nature of that duty? And how do we enforce the execution of said duty? What are the consequences of it not being performed?

You can't really say someone has a "duty" without also implying that they have a "responsibility," and thus liability if they fail to execute those duties properly. I don't see how this fits at all for a volunteer. Very few people are going to volunteer for no pay if they're taking on a risk of liability.

Maybe you mean a civic duty? That would make somewhat more sense, but the problem is that there’s no objective standard against which to test performance. It’s completely subjective and will be forever argued—much like this thread. :-)

If the maintainer is trying to write something RFC-compliant, and someone reports a violation of the RFC, it sure seems reasonable for the maintainer to want to track that.

If they don't want to, that's certainly their right, but it also tells us something about that project.

Someone reporting an RFC violation doesn't automatically mean there is actually an RFC violation. That's why they are asking for a minimal repro, not a dump of the reporter's stream of consciousness. If your teammate at work. came to you and dumped something like this on your desk, how would you react?

Presumably, the maintainer wants the best for the product and its users. So they have a definite interest in documenting a todo list.

Presumably, the user wants the best for the product and their ability to use the product. So they have a definite interest in documenting a todo list.

It doesn't make sense for the two to be at war with each other. It is no big deal for the maintainer to ask a favor. It's not too big of a deal for the user to decline. There's no need to attack.

I have often dropped a note to the maintainer of a project I bumped into. I'm sure they would prefer a bug report in their official forge. But I don't really use their software except for this one time. I'm not willing to jump through the hoops to create an account in yet another SaaS just to file this one report. Just dropping them an email was a courtesy. But often they don't interpret it that way. I'm perfectly un-insulted if they just delete my note and never "fix" the issue because it didn't come through proper channels.

No attacks. No war. Just well wishes. But I might very likely avoid the product if I'm ever back in those woods. Not out of anger or retribution. Just because I'll remember that the product had at least one sharp edge for my use case and the maintainer was a bit overwhelmed by the weight of supporting my niche use case. That doesn't make the maintainer a bad person or even a bad maintainer.

Because they're the ones asking for the administrative burden of refiling a basic RFC violation bug?

I can see both ways here.

If the maintainer just opens the concise bug report they want (RFC .... Section ... If TLS1.3 is negotiated and client sends session id, server must send cipherchangespec), they have what they want and can move on with their life.

However, if the maintainer can get the reporter to do it, the reporter has become a better reporter and the world has become a better place.

IMHO, the original bug report was pretty out there. Asking a library developer to debug a client they don't use with a sever they didn't write either is pretty demanding. I know openssl has a minimal server, I expect woflssl does too? that would be easier to debug.

Actually, on re-reading the original report, the reporter links to a discussion where they have all the RFC references. Had the reporter summarized that to begin with, rather than suggesting a whole lot of other stuff (like a different wolfssl issue that has to be completely unrelated), I think the issue would have gone better.

I will further add that putting a MUST in an appendix seems kind of poor editing. It should have been noted in section 4.1.2 and/or 4.1.3 that a non-empty legacy_session_id indicates that the server MUST send a cipher change spec. It's not totally obvious, but if the client requests middlebox compatability, the RFC says the server MUST do it. If the client doesn't request it by sending a legacy session id, the server can still send a superfluous change cipher spec message if it wants, although I don't know if it will help without the session id.

> The maintainer should just open a new issue for RFC compliance himself since that's a pretty big issue and he obviously thinks OP spams too much.

Reading the issue tracker, why would he do that unless he could repro?

> Hi @feld , I can't really tell if this is related to the ticket that you pointed out. I'll be helping you with this issue as well as looking into the other ticket. Can you give me step by step instructions on how to reproduce what you are seeing? Please note that I have limitted experience with HAProxy and Erlang.

> ...

> I've successfully connected to the server with the examples/client/client and I cannot reproduce what you are seeing. I've built with both WOLFSSL_TLS13_MIDDLEBOX_COMPAT defined and undefined.

He only gets a reply six months later!

This, I feel, clearly shows Feld's intentions - he wasn't interested in agetting it fixed, it was not a bug for him, but he was interested in spreading the word about it. i.e. To me, anyway, it looks like Feld is more interested in writing outrage-bait than getting a working product.

I've used WolfSSL in anger and the experience was much better than OpenSSL and AWS-lc.

Looking at the ticket itself, I consider the responses from the dev team to be pretty good support - better than some paid products I have used.

The blog-poster wasn't happy with the issue being closed, so somehow I doubt that opening a new issue and referencing this one would've yielded a different result from what we got now.

This issue has a similar conversational rhythm that led to the AI agent hit piece that was trending yesterday:

https://theshamblog.com/an-ai-agent-published-a-hit-piece-on...

The OPs blog post also reeks of a similar style to the hit piece.

Given the large delay between the initial report and further responses by the user `feld`, I wonder if an OpenClaw agent was given free reign to try to clear up outstanding issues in some project, including handling the communication with the project maintainers?

Maybe I am getting too paranoid..

A reasonable reply indeed from the maintainer, this happens a lot where you think together in an issue and identify whats really wrong near the end. Only then is one able to articulate an issue in a helpful, concise way. Perhaps GH could add a feature to facilitate this pattern.

Worse yet, despite publishing seventeen blog posts between filing the issue and finally responding to it, he has the gall to open with "Sorry I missed your replies (life gets busy)".

I was reading through the complete issue thread and I have to say I probably would side with the wolfSSL maintainers in part but they could have handled it in a nicer way.

"Anthu" only responded with this after "feld" asked why the issue was closed by them, and only then the response you mentioned was written.

"Anthu" could have simply asked before closing the issue and the reporter would have been fine. Like, say "So, this issue meanwhile evolved into RFC compliance and got a bit off track in my opinion. Can you please open up a separate issue for this so we can get this fixed in a more focused manner? That would be very helpful for our workflow. If not, I would open up an issue and reference this one if that's okay with you."

My point is that feld felt a little ignored in their problem, and the support role could have handled it a little nicer. I get that maintainer time is limited, but I would probably recommend an issue template for these matters where there's checkboxes in them like "keep it short, keep it reproducible" and maybe a separate issue template and tag for RFC matters.

On the other hand, "feld"'s blog post reaction was also quite trigger happy and in part in bad faith. They could've communicated the same things in a "non rage mode" after things have calmed down a bit.

The author has spent a lot of time on this as well. I can see both sides. From the author's perspective their focus is their product/system. Any extra time they spend is not contributing to that. They've already spent a fair amount of time helping root cause the issue and from their perspective once it's clear what the issue is they're done. The author also seems to work on open source. In this case they are the customer of a product, granted an open source one, and they've helped the vendor (the maintainer) figure out something is broken. Their expectation is that the vendor takes things on from there and doesn't put up some bureaucracy.

That said ofcourse you paid nothing for this and you should expect nothing but the OSS project also has no expectations that their customers support them if those customers aren't getting their expectations met. In today's world one unhappy customer can give you a pretty bad rep, as is happening here. Now if you don't care then you don't care. But the argument that because your product is "free" then your customers have no voice doesn't sound that great either.

Everyone seems to be pointing how the author disappeared and came back much later. Well, they disappeared because it wasn't a problem or they've worked around it, and came back when they hit the problem again. Just like the maintainer doesn't work for the author the author doesn't work for the maintainer either.

It's also true the ticket now has a lot of history, but the original bug is still the same bug, it's just that now it has been root caused? The maintainer's response of now that you've found a setting that works around the issue you're good and we can close this also is a bit off. And sure, they don't work for anyone so they're welcome to do whatever they want.

As isn't uncommon when two humans communicate online there is some miscommunication here. But you can argue either way. Not being an open source maintainer I don't know what the "protocol" here is but the few times I've filed bugs against an open source product I did personally put in the extra mile to make them actionable. But in my day job I have to deal with all sorts of bug reports and chasing them down to a resolution is part of what makes the product I work on a better one. And yes, I get paid to do that ;)

What tis missing there is a support team and maybe a difference between a customer (user) facing support system and a big tracker.

Support guides the user through the discovery process, which can be messy and go circles, and the result of that is a big which is actionable by a developer.

The blog author seems like a real piece of work. He ghosts the WolfSSL maintainer for over 160 days and when asked to open a new, more specific issue, he instead chooses to write a blog post denigrating the project. The WolfSSL maintainer was nothing but courteous and helpful throughout the entire exchange.

>...they aren't really interested in RFC compliance.

Yeah, well "feld" can't claim to be "interested in RFC compliance" either when he ghosts the issue for months and chooses to write blog posts instead of opening a new issue. Good grief.

If this is what the FreeBSD community is like, I want nothing to do with them.

I don't think it's fair to judge the whole FreeBSD community by one person.

Seriously, where the hell did that come from?

Where did I judge the FreeBSD community?

Probably where you said:

> If this is what the FreeBSD community is like, I want nothing to do with them.

>If

“If you break the law, then you go to jail” is not “you broke the law, you are going to jail”. I didn’t judge the entire FreeBSD community based on this blog post.

Fun game, let's keep playing.

Where did they say you did?

I'm not "playing a game". "Feld" purports to be a FreeBSD ports committer. Someone with commit rights on a major project would know how to properly file issues and work with other maintainers. But "feld" doesn't seem to know how to do that. Perhaps "feld" had a bad day, or maybe him and the rest of the FreeBSD ports contributors/maintainers just operate in this way, I don't know.

>Where did they say you did?

They said it in the part where you got all confused and responded to me.

You're not playing a game? Could have fooled me.

What you originally responded to above:

> I don't think it's fair to judge the whole FreeBSD community by one person.

So they just like, gave us a fun fact, right?

Conversely, I was also apparently just speculating:

> Probably where you said

So many things are possible when we don't want to be found wrong. Including pretending that figurative speech only exists when it's convenient.

Otherwise, I really don't see what would be so hard in understanding why throwing an "if" at the start would still lead to people taking what you said the way they did.

For the record, contrary to your assertion, even your example is affected by this:

> If you break the law, then you go to jail

If you posted this (and even only just this) under a random thread, people would think you're accusing someone (whoever the given thread is most about) to be somehow guilty of some untold crimes and/or that some chatbot got loose. I hope you can appreciate how people would be absolutely correct to think that way.

Who said you did?

Regarding HAProxy, they ended up using AWS-LC in their new Debian/Ubuntu “performance” packages: https://www.haproxy.com/blog/fresh-from-aws-reinvent-superch...

> Asking me to open a new issue to discuss this behavior instead of it being a high priority for them to open up a new issue internally to fix this is odd. I'm not here to do their homework for them.

Why are people so entitled? How much is the author paying WolfSSL to make demands of them?

> Currently I've only identified one victim of this decision, but there's bound to be more out there.

Oh yes, he has become a victim of using a FOSS library.

Neither person is entitled to the work of the other and neither wants to do the work which seems to be how we ended up here. The author can't make demands of the project and so wrote a blog post warning others that it's not production ready and you'll have broken software if you use it. Their conclusion isn't that they must fix it but that you should use a more mature library.

Two adults both defected in the social prisoner's dilemma and so here we are. Both individuals believing to have done free labor for the other and that they should be grateful.

> Oh yes, he has become a victim of using a FOSS library.

many such cases

The "victim" was the Elixir or Erlang library, not himself. To be clear.

I don’t think it was clear, but thanks for the insight.

Was WolfSSL forced upon Elixir or Erlang? Did they purchase it and received a defective product? Are they held hostage by WolfSSL’s decisions? Are they not allowed to modify WolfSSL as needed themselves?

I fail to see any victims beyond perhaps the WolfSSL maintainers for having to suffer such entitlement.

The lib is (AIUI) a client library, meaning that it will fail to communicate with an HTTPS-enabled, WolfSSL) server, ostensibly because the WolfSSL isn’t SOEC-compliant.

You explicitly modeled the situation so that no victims can exist, so please do spare people from the autofellatious poetic questions and remarks about how you fail to see any victims.

> Was WolfSSL forced upon Elixir or Erlang?

Yes actually, and upon others, that's how computer networking works. Did you read the blogpost by the way? Even just the beginning? Really doesn't seem like it.

Hint: there's a reason the word "middlebox" is mentioned 16 times in there, and that the word "server" is mentioned another 6 more.

> Did they purchase it and received a defective product?

They did not purchase a copy, WolfSSL distributed them one for free. The blogpost author did identify the product as defective however, as it allows for and defaults to spec-noncompliant behavior. It stands to reason that this then affects WolfSSL's paying customers (and their downstream customers) too, who might be unknowingly operating or interacting with spec-noncompliant services as a result.

Will people need to read out the whole article for you?

> Are they held hostage by WolfSSL’s decisions?

Yes, and so are others, that's (still) how computer networking works.

> Are they not allowed to modify WolfSSL as needed themselves?

What would they do with it? Put it on a USB stick and stick it up their ass?

> for having to suffer such entitlement

Are you really one to take issue with another person's behavior after this power tantrum?

BearSSL by Thomas Pornin is always worth checking in on, not sure what the current status is but looks like it received a commit last year.

[1] https://bearssl.org

BearSSL is really cool, but it claims beta quality with the latest release in 2018, doesn't support TLS 1.3, and hasn't seen meaningful development in years. It's averaging about 1 commit per year recently, and they're not big ones.

Where is Bellard when we need him?

Most relevantly here, selling a commercial implementation of ASN.1: https://bellard.org/ffasn1/.

We need something with TLS in the name for the next one so people stop getting confused.

MbedTLS[1] got your back!

[1]: https://www.trustedfirmware.org/projects/mbed-tls/

That's being used by Dillo and it's working really well even on legacy computers.

rustls is there. It has TLS in the name, it is good and there is a C FFI wrapper.

Rustls still outsources cryptographic primitives. I believe the currently supported providers of those are… drumroll… AWS-LC and Ring. The latter is a fork of BoringSSL. The article describes AWS-LC and BoringSSL as "Googled and Amazoned to death; they don't care about anyone but their own use cases".

The state of things sucks :-(

there is https://github.com/RustCrypto/rustls-rustcrypto fwiw

It's a great effort, but it's far from usable:

> USE THIS AT YOUR OWN RISK! DO NOT USE THIS IN PRODUCTION

The author also doesn't specify what that even means and what problems it causes

The primitives aren't a problem. You can't write them in any vaguely modern high level language. And when I say "High level" I mean that the way K&R does when they describe their new C programming language as high level. The reason you can't write cryptographic primitives in a high level language is that optimising compilers love clever tricks which offer data dependent performance, across every layer of their design - but in cryptography we want constant execution time regardless of either the plaintext or keys used.

The problem with OpenSSL isn't these cryptographic primitives, that's why you will see basically the same primitives re-used in lots of different places. It's like finding out that the guy who was just arrested for murder also eats pizza. Yeah, people do that. The problem wasn't the pizza, it was the murder. OpenSSL's implementation of the AES cipher isn't broken, the problem is elsewhere.

What? Ring is not even close to a fork of BoringSSL; it merely borrows subroutines from BoringSSL.

Ok, maybe not a fork outright. But the project description says: Most of the C and assembly language code in ring comes from BoringSSL.

That's the proper way to use OpenSSL and derivatives. Their C and assembly code for crypto primatives is good.

Protocol code and x.509 certficate handling will probably be better written in another language.

You might like https://github.com/ctz/graviola/

Also, even if rustls is using aws-lc-rs, you still get the TLS parts from the rustls project, and aws-lc-rs is just lower-level crypto. That means there's less places for Amazon to say no; they either implement an algorithm or don't.

rustls doesn't have its own implementation of cryptography, you have to choose a provider like openssl or aws lc

Or rustcrypto. Rustls is a TLS layer that can wrap any cryptography layer providing the necessary primitives.

There is a rustls side project called Graviola that's building a fast crypto provider in Rust+ASM. It's taken an interesting approach: starting with an assembly library that's been formally proven correct, and then programmatically translating that into Rust with inline assembly that's easy to build with Rust tooling.

A c wrapper to rust feels like we've gone full circle

That would be amazing and really cement the proven value of Rust.

There's even a project for a deliberately OpenSSL drop-in compatible Rustls backed library. It is intended for specific projects because OpenSSL is sprawling and they don't implement most if it, but in principle if you use the same parts of OpenSSL your C likely works with this safer + faster alternative today, why not recommend it to your users.

https://github.com/rustls/rustls-openssl-compat

You're obviously looking for lastLs.

But then how will we spot the pedants.

If you change your software to comply with "middleboxes" that don't follow standards, then you're admitting your own software is faulty, not theirs. In this case, though, the TLS v1.3 standard actually carved out a portion of the standard itself just to comply with shitty middleware. You know what that says to me? Standards are pointless. Just make a middlebox, make it do whatever the hell you want, and everyone else will bend over to support you.

This is yet one more reason why we need software building codes and regulations. If software people are unwilling to protect their own standards, the government should. It might fix the 20-year mistake of allowing "the web" to become a defacto network transport layer and application platform.

No. This just underscored that if you don't encrypt stuff them idiots will break it. Notice that they didn't break any parts of TLS 1.2 which were encrypted, everything they broke is the unencrypted stuff, and so by encrypting more stuff (everything except client hello) in TLS 1.3 we improved that, and then by encrypting even more stuff in ECH (Encrypted Client Hello) we expect to improve it again.

Government regulation is good in that it can work, but it's terrible in that almost every other choice would be better if it works. For TLS 1.3 we made choices which work, if we'd waited for your hypothetical government intervention we'd still be using TLS 1.2 and Trump would presumably be collecting an inaugural "Super good Bank Encryption Champion" trophy from EDCO or somebody who fought against TLS 1.3 because it meant they'd have to actually do a good job.

Usability-wise (I do not need many features or compliance for FIPS) I have been happy with Botan: https://botan.randombit.net/

Can confirm, used Botan in the past and I didn't curse at it a lot. Certainly less than OpenSSL.

> I have been happy with Botan

Botan is under rated.

It is nowhere as optimized as OpenSSL but its APIs are one order of magnitude better to use.

The team behind also demonstrated a pretty serious handling of non-trivial issues like timing attacks.

What really gets me is the commenter at the end of the GH issue lecturing a maintainer on policy in their own tracker.

Many people and projects have tried to ditch OpenSSL in favor of LibreSSL, WolfSSL, MbedTLS, etc, but by now many have returned to OpenSSL. The IQ curve meme with "just use OpenSSL" applies.

I don't see how OpenSSL can recover from it's 3.0 disaster. They would basically have to write off the past few years of development work and start over from version 1.1.1

I have systematically and successfully banned OpenSSL across all of my Rust projects. Sure, RusTLS shares a few C crypto primitives with OpenSSL forks. But I've never been happier with the overall library.

Rustls is great.

I think it's also worth pointing out that https://pkg.go.dev/crypto/tls has no OpenSSL in it.

Nothing wrong with LibreSSL, really.

If it's good enough for openbsd, it's good enough for you as well.

Particularly, they put in a lot of work on making it a drop-in replacement for openssl, and in making the portable version work well in many platforms.

Only for distributions to fail to take the sorely needed step of actually making the switch.

"OpenBSD was probably right. We just need to focus on LibreSSL and forget about all of these other libraries."

Having compiled many of the popular SSL libraries as an end user, on underpowered computers, IMHO LibreSSL has the best compilation process, e.g., least complex, fastest

The library doesn't have all the features of the others but being able to compile it relatively quickly and easily IMHO is itself a "feature"

WolfSSL has many, many options. Accepting the defaults is not sufficient IME.^1 According to the cited HAProxy blog post, AWS-LC is perhaps the fastest SSL library. But Amazon "overlooked" a simple CMake option that actually made it slower than WolfSSL

To summarise, (a) in addition to library "features" I think the compilation process is also important, (b) IME getting what one wants from the various SSL libraries, if even possible, is needlessly complex and (c) FWIW, LibreSSL has (IMO) the least complicated and fastest compilation process

1. It seems like the author did not want to spend the time to learn about all the options. For the end user (cf. "developer") this make sense. As the HAProxy blog post suggests, the SSL libraries that are controlled by people who work for advertising companies, e-commerce companies and CDN companies are naturally going to put their own interests first. Those interests may not always align with the end user's interests

There’s always rustls.

FIPS compliant?

It is if you use the FIPS compliance feature - then you also depend on aws-lc, but only for the crypto primitives.

The state of things sucks :-(

From safety point of view that's actually good enough for "perfect is the enemy of good" to apply here.

Cryptographic primitives are much much safer in C (and assembly) than protocol handling, certificates etc.

They are basically just "fixed size data block in, fixed size data block out". You can't overflow a buffer, you can't use-after-free etc, you can't confuse inner protocol serialization semantics with outer protocol serialization semantics, you can't confuse a state machine, you can't have a concurrency bug[1] etc.

C memory safety vulnerabilities arise from trying to handle these dynamic things which rustls fixes.

(Also, there are third party crypto providers implemented in Rust)

[1] from memory safety pov; for side channels rust doesn't have advantages anyway

My point is that the article this thread is attached to starts out with how BoringSSL and AWS-LC won't cut it. And when rustls is suggested as an alternative, it's important to point out that it requires precisely those two (either one of them).

The article is about TLS. The arguments against those libs don't apply if using them just for the low level crypto algorithms. (Also of course rustls can use other crypto providers besides those)

Then I'm mistaken. Thanks for clarifying.

Go can create C ABI shared libraries, I think OpenSSL-compatible C bindings to Go's crypto/tls would be a really interesting option.

Do you want garbage collection in your SSL?

Better than no memory safety, sure. Also a kernel should be memory safe, so garbage collected.

Garbage collection is not required for memory safety.

Languages that have garbage collection are not all memory safe.

Where do you see the problems? Because of memory that was not cleaned up and leaks secrets?

There the new runtime/secret could help.

Clearing of the secrets is a separate issue from memory allocation mechanism. It must be done all the way from the encryption layer to the program to avoid the leaks.

This is typically not done, only certain parts such as handling of the crypto keys. That's because it's pervasive and requires reworking everything with that in mind (TLS library, web framework, application).

On the other hand the centralization and global usage of GC in the process allows to modify it to always zero out the memory that it deallocated and to do GC at regular intervals so it can have advantage here (it's very easy to inadvertly leak the secrets to some string).

There is also s2n-tls. I'm working on a GSS-based interface to TLS, much like SChannel is an SSPI-based interface to TLS.

> Last updated on 2026-12-13

Yeah, no, I can't find a way to read this in which it's not in the future.

Never wondered why the camera app in your phone also uses this convention for filenames?

What convention? Is the problem here that you didn't realise this claimed to have been written either in December 2026 (ie months in our future) or, if you're willing to stomach a weird YYYY-DD-MM formatting - a mythical thirteenth month of the year?

The actual article was, eventually, updated to give a real date: Last updated on 2026-02-16

But originally it said 2026-12-13 and that's why I commented.

I am an idiot.

I did not notice the date was in the future

Dillo uses mbedTLS, it's fine.

Now what? BearSSL.

MbedTLS

I don't know if it's still the case, but MbetTLS used to change its API as often as I change my socks.

NanoSSL by DigiCert https://dev.digicert.com/trustcore-sdk/nanossl.html

It's opensource -> https://github.com/digicert/trustcore

AGPLv3 so not exactly a drop in replacement, license-wise.

Reply on news.ycombinator.com