:: Re: [Libbitcoin] libsecp256k1 - ini…
Página Principal
Delete this message
Reply to this message
Autor: Eric Voskuil
Data:  
Para: William Swanson
CC: libbitcoin@lists.dyne.org
Assunto: Re: [Libbitcoin] libsecp256k1 - initial integration
This is great William.

I recently completed GMP and Secp256k1 (both with GMP and with OpenSSL), all Secp256k1-GMP tests are passing.

So at this point all of the dependencies necessary to support all libbitcoin libs on Windows (and in Visual Studio) are now packaged and deployed to NuGet.

http://www.nuget.org/packages?q=evoskuil

Also all libbitcoin libs are building properly in VS and will be ready for packaging once the transition from OpenSSL is complete.

e

> On Apr 24, 2014, at 2:46 AM, William Swanson <swansontec@???> wrote:
>
> Hello,
> I have just pushed a "libsecp256k1" branch to both libbitcoin and
> libwallet. This branch makes a good start at integrating libsecp256k1
> as an EC crypto solution in place of OpenSSL.
>
> There is a new source file, ec_keys, which contains the libsecp256k1
> wrapper. This wrapper provides a generic interface to EC crypto stuff,
> and doesn't expose any details of the underlying library. Amir and I
> spent a while talking about how this interface should look, both in
> terms of high-level concept and in terms of little details like proper
> whitespacing. If we change our minds on our dependencies, this wrapper
> could just as easily be reimplemented using OpenSSL.
>
> This wrapper obsoletes the elliptic_curve_key class, which we can
> eventually remove.
>
> With this wrapper in place, I went ahead and ported my bip32 hd_keys
> classes to the new system. Now there are zero references to OpenSSL in
> the HD wallet code, not even for bignums. All unit tests pass as
> before.
>
> There is still a bit of work to do going forward. The secp256k1 and
> gmp libraries need to be packaged for Windows, and the rest of the
> codebase needs to be ported to the new wrapper class. The wrapper is
> also missing the ability to generate random nonces for it's EC
> signatures. This is a *CRITICAL* security hole, but fortunately, it's
> an easy fix.
>
> I will be rebasing the libsecp256k1 branches from time to time as I
> continue to work on this stuff.
>
> -William
> _______________________________________________
> Libbitcoin mailing list
> Libbitcoin@???
> https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/libbitcoin