On 2011 December 16 Friday, Rick Wesson wrote:

> I believe that any URI scheme will still leverage DNS and inherit any
> base issues you would have with TXT records. I suggest looking at DANE

HTTPS takes care of that.

> and reviewing their work on hardening certificate (x.509)
> infrastructure as your HTTPS scheme will inherit the issues we
> currently experience with CAs getting p0wned.

This is the only real problem with HTTPS: we would be centralising part of our 
otherwise decentralised system.  CAs are certainly a risk.

However, trust is needed somewhere in the communication.  There is no way to 
securely communicate between A and B without the use of some previously 
trusted secure channel -- in Joe Sixpack's case it's by assuming that the 
browser he downloaded came with an untainted CA list, and that the CAs are 
trustworthy.  Neither of which is guaranteed.  Until and unless we get PGP 
support in browsers, CAs are all that we have.

Worrying about CAs misses the point anyway; if we're being that paranoid -- 
how did A tell B the appropriate alias to use for a lookup?  Was that channel 
secure too?  I could set up a MITM server that simply looks for the alias 
"RICKWESSON@bitcoinaliases.org" and rewrites it to 
"ANDYPARKINS@bitcoinaliases.org".  When the answer to that problem is HTTPS 
(or some other system that requires a previously authorised secure channel for 
transfer of trust), then we're back where we started, and HTTPS is acceptable.



Andy

-- 
Dr Andy Parkins
andyparkins@gmail.com