[readme] Add section Bindings #381

I think section in readme about bindings is better than folding bindings code in this repository (like with Java).

Related issue: #255

Bindings authors: @toxeus @xenog @wanderer @afk11 @ludbb @DavidEGrayson @wink @dpc

I disagree that we should be recommending bindings that we have not reviewed. Previously some bindings have compromised the security of the library. (e.g. the go bindings silently using insecure nonce generation).

@apoelstra : What's the story with https://github.com/dpc/bitcoin-secp256k1-rs ? Did you supersede it with some other repository?

@gmaxwell there no words about recommended, but maybe it make sense add note that bindings wasn't reviewed? OOI, whom you had in mind under we?

Directing people to things which may be insecure like that is just 'laundering' when we know the note would just be ignored-- it might be necessary but it's not sufficient: Assuming one of the bindings is incorrect (or even malicious) listing it with a note would be a boobytrap and incompatible with our design principle of being hard to misuse.

Bindings should be subject to peer review by cryptographers; same as the extern C interface of libsecp256k1, since it's functionally the same when used in different environments. The review will be easier when we finish documenting our design assumptions and objectives; since then the bindings could be checked against satisfying those properties.

@gmaxwell : People will just google "bitcoin bindings in X" or use search function in the package manager of the language of their choice etc.

IMO having a "big red warning" on the list here, and requiring projects listed in the list to have one in the README, is a win, as it has some chance of informing the potential users.

Oh, I see you found my Ruby bindings. I haven't updated them since April 2015 so I suspect that they don't work with the latest version of the library. I was going to wait for you guys to release version 1.0.0 of the library before I work on that again. Once everything is polished up, I'd be happy to help cryptographers review it.

We can include a warning without recommending any bindings: "Note that bindings to other languages may undermine the security of the library and have done so in the past. Great care should be taken when using bindings found on the internet as they may not be suitable for your purposes."

(e.g. the go bindings silently using insecure nonce generation)

Are you referring to this?

@dpc Sorry I missed this message. Yes, the rust libsecp bindings are now at https://github.com/apoelstra/rust-secp256k1

I think if you decide to add a list of 'recommended' bindings then there should be a clear list of review criteria the projects have to adhere to, and a list of tests that it has to pass in the bound language.

But to take the task of 'vetting' projects is quite a responsibility and I can understand if @gmaxwell and al aren't willing to include that in their scope.

This may be better for a wiki article or such, with the appropriate warnings.

On the other hand, good bindings are important for adoption of the library. Putting the bar too high may result in people just ignoring advice, or rolling their own without cooperation and review of others.

I'll glad if somebody will add this list into wiki.