./bitcoind send genjix@foo.org 0.1 #169

OK, lets try this in smaller more informative commits.

Here's the first of a sequence of pull requests. This one only adds the ability to send to a named@server

It expects back from the server a JSON with an address key or one with "error"/"errcode" in which case it throws an error in RPC. There is an example implementation of the server code in contrib/ns/ which can be installed.

12:14 < sipa> genjix: i mean even further, it should be clear that you're able to have a name-lookup service without any ability to log in on a server, fiddle with public keys, know that it uses rsa signatures, ...

FYI there are several parts to my branch:

• It's a simple name lookup from a webserver.*
• An example PHP implementation that returns the addresses.
• An RSA keypair class which supports displaying a public key in PEM format.
• An example PHP implementation for a service which allows you to update your public key using an OpenID login (supports Facebook, Google, Yahoo, Wordpress, Launchpad .etc)
• A name lookup update which takes your nickname+address+a timestamp in the POST request and signs it so the server can verify it's you.
• An example PHP name lookup which allows you to set your address.
• This latest pull request is only for this first part.

more info here: http://www.bitcoin.org/smf/index.php?topic=5938.msg91331

this patch also adds libcurl as a required library (good IMHO, and useful for more things)

I do not have any specific technical objection, but I just don't think this belongs in bitcoind. It does not seem like a feature that would be used by more than a few people, especially when considering #174 may get merged (I think it will, anyway).

Why? My branch has much more technical depth than that other one. It's also written cleaner and much easier to extend. The other branch is simply looking up an address, whereas I have signing the requests and an example server for setting your address written in PHP.

IMO there is not enough security in something like this. It suffers from the same problem that IP-based transactions do, namely that it is subject to MITM attacks. This approach is further subject to DNS poisoning attacks.

If this is going to work securely, the request must be made over HTTPS, and the server's certificate must be verified somehow.

It's already using HTTPS.

It's already using HTTPS.

Are you sure about that? I see this:

string strRequestUrl = strDomain + "/getaddress/?nickname=" + pszEncodedNick;

And strDomain does not start with https://. I don't see any curl option forcing HTTPS either.

I agree that this sort of endpoint-to-address discovery should not be in bitcoind. There are several big problems:

1. Lack of widely-deployed DNSSEC means this is hard to secure
2. It doesn't use SSL, and probably shouldn't use HTTP(S) at all. DNS SRV or TXT records would be better for such a scheme.
3. The URL scheme is half-baked. "getaddress" says nothing about bitcoin and could easily conflict with other uses of a website. Again, I don't think that HTTP is the best transport for this sort of discovery.

No consensus that this is the right way to to; I'm going to close.