> On Oct 24, 2017, at 5:26 PM, Michael Richardson <mcr+i...@sandelman.ca> wrote:
>
> In the browser space there has been pushback against including the trust
> anchors in the Server->Browser direction, including Google's Chrome browser
> complaining about unnecessary certificates, and TLS scanners.
> I understand that some of this is the result of some client libraries that
> could be confused (due to bugs) into validating a bogus chain if there was a
> self-signed certificate in the certificates sent from the server.
>
> What's unclear to me if there is any kind of specification that we would be
> violating if we state that we want the full chain in the Client's Certificate
> extension.
Full chains are just fine. Indeed per RFC7671 with DANE-TA(2) the server
MUST present a full chain (including the root CA certificate) to the client
when the server's TLSA record is associated with the trust anchor certificate.
If you have a legitimate use case in which the relying party may not have
a copy of a root CA, but can validate it if received from the peer, then
requiring the transmission of root CAs is fine and natural.
--
Viktor.
_______________________________________________
TLS mailing list
TLS@ietf.org
https://www.ietf.org/mailman/listinfo/tls
