Re: [TLS] TLS specification clarification in case of client authentication: different CA with DN different only in case

[Geoffrey Keating <[e-mail]>]

devzero2000 <[e-mail]> writes:

> Hello everyone
> 
> >From the tls 1.2 specification, speaking of client authentication,
> https://tools.ietf.org/html/rfc5246#section-7.4.4 par 7.4.4 (but it is the
> same for the last tls draft 1.3 par. 4.2.4.)
> 
> when he says:
> 
> certificate_authorities
>       A list of the distinguished names [X501] of acceptable
>       certificate_authorities, represented in DER-encoded format.
> 
> What would be the right behavior if the server has the certificates of two
> different CAs (different subject key info, public key parameter) but whose
> subject DN differs only for the case (for example
> something like this
> 
> Subject: /C=US/ST=California/L=Mountain View/O=XXX Inc/CN=mail.xxx.com
> 
> and
> 
> 
> Subject: /C=US/ST=California/L=mountain View/O=XXX Inc/CN=mail.xxx.com

These are the same distinguished name under RFC 5280 section 7.1,
although in practice implementations may treat them as different, most
notably under the older RFC 3280 rules.  I believe the correct
behaviour is to Not Do That---do not generate certificates which have
distinguished names that match under RFC 5280 and are not
byte-for-byte identical in DER format, if you must do that make sure
they are not valid at the same time, and if you must do that, try to
ensure no piece of software is aware of both of them, and if you must
do that, don't be surprised if the behaviour is inconsistent and
especially don't be surprised if the LDAP StringPrep rules are not
implemented correctly or at all.

And if you value your sanity, don't rely on anything that might change
if the Unicode standard is revised.

However, the TLS specification doesn't say that the list must contain
each DN only once.  So in this situation I would suggest the software
should list both.  Indeed I would recommend listing every distinct DER
representation that's present in any acceptable certificate.