David Benjamin

I look forward to all the latent X.509 bugs around 2050, as we rollover from UTCTime to GeneralizedTime. Did you think we were done with two-digit years after Y2K? Well, X.509 has some news for you!

Hey @thedarktangent, defcon.org is TLS 1.3 intolerant and will break in Chrome & Firefox soon. Are you running a prerelease OpenSSL 1.1.1? They had a version negotiation bug: github.com/openssl/openssl/i…

TLS 1.3 (draft 23) is rolling out with Chrome 65! Let's hope it sticks. Third time's the charm?

Anyone have a contact on Fortinet's firewall product? Could you have them mail davidben at google? (This is about TLS 1.3 compatibility.)

Use Chrome beta? If you're up to date, about half of you have TLS 1.3 (draft 18). Enjoy!

Dear servers, it is *not possible* to implement TLS 1.2 without SHA-2. Why are so many of you signing TLS 1.2 ServerKeyExchanges with SHA-1?

NSS can now run against BoringSSL's test suite! It's so great to see our tests be useful for a second TLS stack. ietf.org/mail-archive/web/tl…

If you try something three times and it does not work, don't do it a fourth time. I'd really like to be done with the TLS version fallback.

RSA-PSS/SHA-512/sLen=hLen is too large for 1024-bit RSA. Test credentials are often 1024-bit. TLS 1.3 is RSA-PSS. This combines hilariously.

Good-bye and good riddance! (At least until TLS 1.3...) It only took, what, 15 years? chromium.googlesource.com/ch…

C, it's 2016. We're really quite confident about the whole two's complement thing now.

I've finally open-sourced my holiday hacking. It's a reversible DER (or BER) pretty-printer! github.com/google/der-ascii

It is a little terrifying when Twitter notices within minutes of my sending an email...

If you value your sanity, don't implement EC private key serialization. It's such a mess. Excess options, conflicting specs, legacy bugs...

Ugh, my Twitter stream is full of bad crypto today. Here, have some good news: boringssl.googlesource.com/b… (Holidays, so not on by default yet.)

Implementing something twice in two different languages doesn't really count as interop testing, but it sure is handy in a pinch.

Okay, Linux and Chrome OS users, your turn! Hopefully I didn't break much. (Linux/CrOS Chrome 46 is BoringSSL, joining Android/Mac/Windows.)

So not only do I have to allow negatives, but also extra leading zeros. It shouldn't be so difficult to encode a DER INTEGER right...

This is your friendly routine reminder that, no, lashing out at folks trying to help you is still not a good way to get your bugs fixed.

Every time I think I've seen the weirdest TLS server bug possible, someone manages to find another even weirder one.