Re: [saag] AD review of draft-iab-crypto-alg-agility-06

[Martin Thomson <[e-mail]>]

On 27 August 2015 at 13:37, Watson Ladd <[e-mail]> wrote:
> At what point can we remove the code from OpenSSL? That's the endpoint
> we should be aiming for.

I expect this to progress roughly as follows (for Firefox, and in general):
1. announcement
2. disable, maybe with a way to manually override
3. disable with no override
4. disable this on the last of the servers
5. remove from the library

You are asking about step 5, which is important, but some of the
intermediate steps have a non-trivial benefit too.

The reason for step 4 is a little ugly.  We run openssl on our
download servers.  We want to offer Firefox downloads to users who
have old and busted browsers.  That means we need to be a little more
promiscuous on those servers than we might otherwise want to be.  That
means that we also don't do anything much of importance on those same
servers.

We want security updates too, so we want new versions of openssl that
will support us.  That means that openssl will likely support old and
crufty stuff for longer than you might think.

We still haven't disabled SSLv2 in NSS, but that's laziness.  The
reason will still have SSLv3 in NSS is in part due to step 4.