5 ms·
I'm torn about this. Brought to you by the person that wrote "Rolling your own crypto" (see earlier discussion linked below), with the motto "Impossible? Like
by FabHK 10y ago
I'm torn about this.
Brought to you by the person that wrote "Rolling your own crypto" (see earlier discussion linked below), with the motto "Impossible? Like that would stop me."
Some fields of knowledge in certain phases benefit from iconoclasm and Sturm & Drang, but is cryptography one of them? On the other hand, many existing implementations are so crufty that a fresh wind might help (if only by pushing more experienced cryptographers to put out better implementations).
At any rate, at least the writer followed through, though I'd not be inclined to use this in production anytime soon.
http://loup-vaillant.fr/articles/rolling-your-own-crypto http://loup-vaillant.fr/articles/rolling-your-own-crypto
https://news.ycombinator.com/item?id=13221923 https://news.ycombinator.com/item?id=13221923
- lvh 10y agoRight there with you. A question that helped answer this for me personally: does this library need to exist given that libsodium already exists? The answer seems "no" to me.
- loup-vaillant 10y agoThe very existence of TweeNaCl indicates that DJB would have answered "yes". And I do have a couple minor advantages over TweetNaCl…
- lvh 10y agoTweetNaCl was created approximately simultaneously (2013) with libsodium.
- loup-vaillant 10y agoHmm, that explains a number of things. Still, I've dug through libsodium, and it's still a bit of a mess —not that I would've done better given the constraints they had.
- VeejayRampay 10y agoThen again, the author worked on the topic, doesn't try to hide the fact that this is a personal project and provides open source, auditable code. As much as I understand that crypto is somehow special and cannot be done "almost right", this is just someone bringing something to the table no matter what. Maybe it will somehow encourage others to try to understand the concepts behind cryptography, which will foster exchange and debate, which is always good.
- loup-vaillant 10y ago> I'd not be inclined to use this in production anytime soon. Indeed you should not. Rolling crypto on your own is too damn dangerous, and that's what I just did. On the other hand, I'm convince there's not much left to do to ensure a sufficient level of confidence, for 3 reasons: 1) The library is small. Almost as small as TweetNaCl itself. 2) With the exception of Poly1305 and curve25519, the primitives chosen here are simple and easy to test. If I didn't royally screw up somewhere (and I need external verification to make sure I didn't), those primitives are all production worthy. 3) curve25519 is is directly from TweetNaCl with very few modifications. The hard part (ensuring there are no rarely occurring limb overflow) has already been done there. The weakest link for now is Poly1305: limb overflows are possible, and I did it from scratch. There's no way I can trust it without external review.