I'm really glad to see this proposal.  We already treat non-DER
signatures as non-standard in btcd and agree that extending them be
illegal as a part of a soft fork is a smart and sane thing to do.

It's also good to see the explicit use of signature parsing since it
matches what we already do as well because we noticed noticed OpenSSL's
notion of big numbers (unsigned) didn't agree with Go's (signed).  By
having the explicit signature scheme and checking clearly called out in
a BIP, it greatly lowers the chances of there being any disagreement
about what is valid or invalid due to an underlying dependency.

+1

On 1/20/2015 6:35 PM, Pieter Wuille wrote:
> Hello everyone,
> 
> We've been aware of the risk of depending on OpenSSL for consensus
> rules for a while, and were trying to get rid of this as part of BIP
> 62 (malleability protection), which was however postponed due to
> unforeseen complexities. The recent evens (see the thread titled
> "OpenSSL 1.0.0p / 1.0.1k incompatible, causes blockchain rejection."
> on this mailing list) have made it clear that the problem is very
> real, however, and I would prefer to have a fundamental solution for
> it sooner rather than later.
> 
> I therefore propose a softfork to make non-DER signatures illegal
> (they've been non-standard since v0.8.0). A draft BIP text can be
> found on:
> 
>     https://gist.github.com/sipa/5d12c343746dad376c80
> 
> The document includes motivation and specification. In addition, an
> implementation (including unit tests derived from the BIP text) can be
> found on:
> 
>     https://github.com/sipa/bitcoin/commit/bipstrictder
> 
> Comments/criticisms are very welcome, but I'd prefer keeping the
> discussion here on the mailinglist (which is more accessible than on
> the gist).
>