Bug #9424: ruby 1.9 & 2.x has insecure SSL/TLS client defaults | Hacker News Reader

Bug #9424: ruby 1.9 & 2.x has insecure SSL/TLS client defaults

bugs.ruby-lang.org

90 pointsmikeevans26 comments

There's been a lot of anger around Twitter on this. I've also seen a lot of people cherry-picking a non-native speaker's words out of context too. Specifically, "Ruby is not a project for security."

That doesn't mean that this bug is not important, or that the Ruby team's decision as it currently stands is a good one. But it's a complex issue.

If someone gave me that sentence, I wouldn't know whether the person that wrote it was a native speaker or not.

Exactly. And it implies something different than what was meant if you read the surrounding context.

"It's a complex issues" == Ruby Security Fails again.

It is a crackers dream that so much Ruby code is being exposed to the web these days. Such low hanging fruit. Even the script kiddies laugh at the ease of compromise.

Sigh.

A throwaway troll account, how novel.

"Ruby is not a project for security."

That's from ruby-core. That's a frightening attitude for a project to take.

Did you actually read the context of that quote, which happens to communicate almost exactly the opposite of what you're inferring (and implying by quoting it out of context)?

From the context, it seems like mame was trying to say that Ruby is not a security-focused project, so the core team has not attracted many volunteers who are familiar with SSL/TLS.

I inferred from this not that Ruby team doesn't care about security, but that they lack the expertise to handle it properly. They're aware of that, and choose to leave these decisions up to the experts.

It's a reasonable position, but, as a user, "We don't know how" doesn't help me any more than "We don't care".

Sure, the context of the thread is that ruby-core is reluctant to start changing OpenSSL defaults because Security is Hard, the ruby team is volunteers who might not have expert-level crypto understanding, and and you risk doing more harm than good if you mess with crypto when you do not know what you're doing, so they'd rather leave OpenSSL's defaults alone, and let OpenSSL, written and reviewed by security focused people, fix any issues (I think this is a fair summary?).

While these are valid concerns, you can't wholly pass the buck to OpenSSL and to people installing a new version of OpenSSL and re-linking ruby against that -- ruby project should should always ship the most secure ruby possible, all the time.

Newer OpenSSL has already changed these defaults, smart people who know what they are doing have already agreed and documented that this is a good idea, and ruby is only hurting their users by dodging responsibility.

edit: spelling.

Yes, I agree - the ruby project should always ship the most secure ruby possible, all the time.

Would you rather have a ruby-core of honest, hardworking developers who prefer to "delegate the task to other people" who are experts in security - because they admit their own weaknesses, or a ruby-core of arrogant developers who overstep their area of expertise just to try and deliver your "most secure ruby possible" - but more than likely worsening security?

It sounds to me like they're doing the most responsible thing they possibly can given the circumstances. If you or anyone else is in a position to do better, then by all means, step up and bless the ruby world with your supreme intellect.

admitting weakness + asking for help != dodging responsibility

Let us present a third possibility: Maybe they shouldn't have shipped a TLS client if they weren't prepared to take responsibility for how it was configured.

True, and by that same logic maybe anyone using ruby's TLS client shouldn't ship anything if they aren't prepared to take responsibility for how it's configured (since the default configuration can be overridden at any point from OpenSSL on up).

Sure, any consumer of this API really should fix it themselves to have a secure configuration. But I submit that basically none of them are. An API where almost everyone who uses it wrong is not a good API. When the consequence of using it wrong is poor security, that's a dangerous API.

Knowingly shipping dangerous APIs is irresponsible.

OpenSSL is a god damned shitshow, no questions from me, it's bad, it's dangerous, it's irresponsible.

But they shipped something based on OpenSSL, and now they're making a deliberate decision not to act to protect their users. That's not cool, and that's unacceptable to me. If I actually used Ruby, this would make me reconsider that.

Importantly, every layer in the stack is responsible for it's own security. A consumer of this API should be making sure that it's optimally configured and configure it differently where it's not. Any project that isn't doing that should have security reports sent to it to tell it to do that and if they refuse they are guilty of the same sort of negligence as ruby core.

However the fact that other people should also be claiming responsibility for their own security does not absolve ruby of it's own responsibilities.

At least they are being honest I suppose. If you want a secure language and ecosystem - don't use Ruby.

Security - they've heard of it, at least now.

A throwaway troll account, how novel.

How about if you aren't a security expert you shouldn't be programming at all because it's all about the weakest link, right? Don't use Java, C, C++ either.

Now it's got deleted. Anyone have a screenshot or cache or something?

Non-SSL expert here and first time poster (not trolling). Python also uses a wrapper for OpenSSL and has similar issues with default settings. Is this problem specific to Ruby or also Python apps as well?

This also applies to Python. For example, Python recently disabled SSLv2. Someone filed a bug and the fix was committed within 3 hours. http://bugs.python.org/issue20207

The SSL2 ClientHello thing is, IIRC, also a compat hack; Firefox used it (at least until recently) when it connected through proxies.

Firefox stopped using it on 2011-08-16. If you've seen Firefox using it recently when connecting through proxies, please let me know.

I'm sure you're right. It was on whatever version of Firefox was on my work laptop the week I learned Golang, since my first project with the language was an HTTPS/SSL proxy. I even griped about it on Twitter (it breaks Golang's crypto/tls), and (after we'd hacked support for it into our local tree) got a link to a patch. That would have been... meh... a year and a half two years ago.

Anyone know of a simple recipe to set up secure defaults?

This is a good place to start on good OpenSSL defaults.

https://www.ssllabs.com/projects/best-practices/index.html

So does PHP, though this is fixed in 5.5 IIRC.

Reply on news.ycombinator.com