What? Come on. Get a grip on reality. The code is some of the most efficient and widely used. I think it's also the only code that implements the entire spec.
It's like those commenting "that skyscraper is garbage" by someone who doesn't even know what you'd need to build a skyscraper, let alone actually tried to build one.
It's like those commenting "that skyscraper is garbage" by someone who doesn't even know what you'd need to build a skyscraper, let alone actually tried to build one.
It's nothing like that at all. A piece of code's popularity is no evidence of its quality. Otherwise we'd never have advanced beyond BASIC.
Implementing an entire spec is nothing to be proud of. Specs are often over-engineered. This is almost certainly the case for TLS. It's 103 pages long: http://tools.ietf.org/html/rfc5246
We should be questioning the foundations we've been relying on until now when a single vulnerability was able to compromise most of the internet, possibly since 2011.
> The code is some of the most efficient and widely used.
Efficient, yes. Throughput in general is not a problem.
Widely used, well... yes to that too. Although in this regard I would attributed a good chunk of it to inertia. When (common) use of cryptography started to take off around the time of the browser wars, there weren't too many options around. Web servers needed a way to serve protected content, and there was OpenSSL.
Not just available, but already supported by language bindings and relatively easy to set up thanks to mod_ssl.
First mover advantage + network effects + the fact that getting crypto right is really hard. If you wanted to unseat OpenSSL, you would need to overcome the fact that despite all the bad rap and the butt-ugly, and the leaky API, it was still getting the most exposure and field testing. Unlike some newcomer library, OpenSSL has the advantage that it is firmly embedded in the software ecosystem, and it has benefited from more than a decade of bugfixes. (I still remember when the timing attack for extracting RSA keys was discovered. The bug was pretty much everywhere, and not just in OpenSSL.)
The widespread use of the library boils down to a simple and womewhat depressing fact: OpenSSL is used everywhere because OpenSSL is used everywhere.
Replacing it with something better will be a long uphill battle. Any newcomer would need an aggressive test suite which covered EVERY known implementation bug. And a sane API. And a project with the political savvy to succesfully lobby for its use through all avenues.
If the skyscraper collapses on me I feel entitled to say that it was garbage even if I'm not capable of building one.
But I am a half decent C programmer and I saw the code that included the heartbleed bug. This bug wasn't caused by a subtle issue. It wasn't even a logical error that a competent person could easily make in a confused moment. This was pure negligence, and negligence is rarely isolated in one part of a codebase.
PHP5 is widely used. The security history of that is less than awesome.
I think OpenSSL is better than I could do or most people could do, but again, it's still at the mercy of C AND a wide range of TLS and other features that seem to multiply like rabbits.
I think we really need a much, much simpler TLS standard going forward.
What? Come on. Get a grip on reality. The code is some of the most efficient and widely used. I think it's also the only code that implements the entire spec.
It's like those commenting "that skyscraper is garbage" by someone who doesn't even know what you'd need to build a skyscraper, let alone actually tried to build one.