Re: [TLS] Generalising DN's to SAN and IAN in TLS1.3?

[Andrei Popov]

Hi Henry,

Extension_data field can be zero-length:
opaque extension_data<0..2^16-1>;

The TLS 1.3 draft specifies matching rules for two extensions: KU and EKU and 
then says:
"Separate specifications may define matching rules for other certificate 
extensions."

Which means that you should be able to specify matching rules e.g. for SAN or 
IAN where zero-length extension_data is interpreted as a wildcard. Getting 
TLS/PKI implementations to implement this new specification is a different 
matter, of course.

Cheers,

Andrei

From: TLS [mailto:tls-boun...@ietf.org] On Behalf Of Henry Story
Sent: Wednesday, March 16, 2016 5:01 AM
To: Eric Rescorla 
Cc:  
Subject: Re: [TLS] Generalising DN's to SAN and IAN in TLS1.3?


On 8 Mar 2016, at 09:30, Eric Rescorla > 
wrote:

CertificateRequest already allows you to pass an arbitrary number of
extensions by OID.

http://tlswg.github.io/tls13-spec/#certificate-request

What more do you think you need?

Perhaps what would be needed in addition would be wildcard support.

Eg, it would be useful to say: Give me certificates that contain an extension 
without
specifying the value of the extension. Eg: give me a certificate that contains a
SAN or IAN, I don't care what value of those are.

Is that allowed? I don't see anything regarding it when reading that section. 
But
I may be missing something.



-Ekr


On Tue, Mar 8, 2016 at 12:22 AM, Henry Story 
> wrote:
Hi,


  I was reading with interest M. Thomson and M. Bishop's
"Reactive Certificate-Based Client Authentication" draft RFC [1].
In the section 2.3 "The CERTIFICATE_REQUEST Frame"

[[
  CA-Count and Certificate-Authorities:  "Certificate-Authorities" is a
  series of distinguished names of acceptable certificate
  authorities, represented in DER-encoded [X690] format.  These
  distinguished names may specify a desired distinguished name for a
  root CA or for a subordinate CA; thus, this message can be used to
  describe known roots as well as a desired authorization space.
  The number of such structures is given by the 16-bit "CA-Count"
  field, which MAY be zero.  If the "CA-Count" field is zero, then
  the client MAY send any certificate that meets the rest of the
  selection criteria in the "CERTIFICATE_REQUEST", unless there is
  some external arrangement to the contrary.
]]

Would it not be possible to extend that so that one could also pass
issuer Alternative Names, and not just DNs?

Using DNs made sense when it was thought that LDAP and X500 would
end up being the protocols for global directories. This turned out
not to be the case. It turned out that DNs were a special case of
what could be termed a URI (even though DNs are not in URI format).
And many (most?) URIs can refer to agents, least but not last
http(s) URLs (See the WebID spec [2] for a nice diagram of how this
works conceptually and the WebID-TLS spec for one way this can be tied
to TLS [3]).

If TLS1.3 could start moving away from sole reliance on DNs this
would open quite a lot of possibilities, including the ability to
build institutional Webs of trust for CAs (ie trust anchors could
list CAs by reference at one or more levels of indirection), and CAs
could also describe themselves at their URI.

Those last two paragraphs are just hints of some possibilities that
could emerge from moving away from DNs that I can think of. Others
members of this group will certainly find other possibilities.

In any case it seems that the time for DNs is passed, and that
one should perhaps move away from reliance on those and generalise
this part of TLS.

Henry



[1] 
https://tools.ietf.org/html/draft-thomson-http2-client-certs-01
[2] 
https://www.w3.org/2005/Incubator/webid/spec/identity/#overview
[3] 

Re: [TLS] Generalising DN's to SAN and IAN in TLS1.3?

[Henry Story]

> On 8 Mar 2016, at 09:30, Eric Rescorla  wrote:
> 
> CertificateRequest already allows you to pass an arbitrary number of
> extensions by OID.
> 
> http://tlswg.github.io/tls13-spec/#certificate-request 
> 
> 
> What more do you think you need?

Perhaps what would be needed in addition would be wildcard support.

Eg, it would be useful to say: Give me certificates that contain an extension 
without
specifying the value of the extension. Eg: give me a certificate that contains 
a 
SAN or IAN, I don't care what value of those are.

Is that allowed? I don't see anything regarding it when reading that section. 
But
I may be missing something.

> 
> -Ekr
> 
> 
> On Tue, Mar 8, 2016 at 12:22 AM, Henry Story  > wrote:
> Hi,
> 
> 
>   I was reading with interest M. Thomson and M. Bishop's
> "Reactive Certificate-Based Client Authentication" draft RFC [1].
> In the section 2.3 "The CERTIFICATE_REQUEST Frame"
> 
> [[
>   CA-Count and Certificate-Authorities:  "Certificate-Authorities" is a
>   series of distinguished names of acceptable certificate
>   authorities, represented in DER-encoded [X690] format.  These
>   distinguished names may specify a desired distinguished name for a
>   root CA or for a subordinate CA; thus, this message can be used to
>   describe known roots as well as a desired authorization space.
>   The number of such structures is given by the 16-bit "CA-Count"
>   field, which MAY be zero.  If the "CA-Count" field is zero, then
>   the client MAY send any certificate that meets the rest of the
>   selection criteria in the "CERTIFICATE_REQUEST", unless there is
>   some external arrangement to the contrary.
> ]]
> 
> Would it not be possible to extend that so that one could also pass
> issuer Alternative Names, and not just DNs?
> 
> Using DNs made sense when it was thought that LDAP and X500 would
> end up being the protocols for global directories. This turned out
> not to be the case. It turned out that DNs were a special case of
> what could be termed a URI (even though DNs are not in URI format).
> And many (most?) URIs can refer to agents, least but not last
> http(s) URLs (See the WebID spec [2] for a nice diagram of how this
> works conceptually and the WebID-TLS spec for one way this can be tied
> to TLS [3]).
> 
> If TLS1.3 could start moving away from sole reliance on DNs this
> would open quite a lot of possibilities, including the ability to
> build institutional Webs of trust for CAs (ie trust anchors could
> list CAs by reference at one or more levels of indirection), and CAs
> could also describe themselves at their URI.
> 
> Those last two paragraphs are just hints of some possibilities that
> could emerge from moving away from DNs that I can think of. Others
> members of this group will certainly find other possibilities.
> 
> In any case it seems that the time for DNs is passed, and that
> one should perhaps move away from reliance on those and generalise
> this part of TLS.
> 
> Henry
> 
> 
> 
> [1] https://tools.ietf.org/html/draft-thomson-http2-client-certs-01 
> 
> [2] https://www.w3.org/2005/Incubator/webid/spec/identity/#overview 
> 
> [3] https://www.w3.org/2005/Incubator/webid/spec/tls/ 
> 
> 
> ___
> TLS mailing list
> TLS@ietf.org 
> https://www.ietf.org/mailman/listinfo/tls 
> 
> 

___
TLS mailing list
TLS@ietf.org
https://www.ietf.org/mailman/listinfo/tls

Re: [TLS] Generalising DN's to SAN and IAN in TLS1.3?

[Henry Story]

> On 8 Mar 2016, at 08:30, Eric Rescorla  wrote:
> 
> CertificateRequest already allows you to pass an arbitrary number of
> extensions by OID.
> 
> http://tlswg.github.io/tls13-spec/#certificate-request 
> 
> 
> What more do you think you need?

If that would allow one to specify that certificates that match a specific
IAN are acceptable and if those get implemented widely, then that's
what I was looking for. 

Thanks. Look forward to that :-)

Henry

> 
> -Ekr
> 
> 
> On Tue, Mar 8, 2016 at 12:22 AM, Henry Story  > wrote:
> Hi,
> 
> 
>   I was reading with interest M. Thomson and M. Bishop's
> "Reactive Certificate-Based Client Authentication" draft RFC [1].
> In the section 2.3 "The CERTIFICATE_REQUEST Frame"
> 
> [[
>   CA-Count and Certificate-Authorities:  "Certificate-Authorities" is a
>   series of distinguished names of acceptable certificate
>   authorities, represented in DER-encoded [X690] format.  These
>   distinguished names may specify a desired distinguished name for a
>   root CA or for a subordinate CA; thus, this message can be used to
>   describe known roots as well as a desired authorization space.
>   The number of such structures is given by the 16-bit "CA-Count"
>   field, which MAY be zero.  If the "CA-Count" field is zero, then
>   the client MAY send any certificate that meets the rest of the
>   selection criteria in the "CERTIFICATE_REQUEST", unless there is
>   some external arrangement to the contrary.
> ]]
> 
> Would it not be possible to extend that so that one could also pass
> issuer Alternative Names, and not just DNs?
> 
> Using DNs made sense when it was thought that LDAP and X500 would
> end up being the protocols for global directories. This turned out
> not to be the case. It turned out that DNs were a special case of
> what could be termed a URI (even though DNs are not in URI format).
> And many (most?) URIs can refer to agents, least but not last
> http(s) URLs (See the WebID spec [2] for a nice diagram of how this
> works conceptually and the WebID-TLS spec for one way this can be tied
> to TLS [3]).
> 
> If TLS1.3 could start moving away from sole reliance on DNs this
> would open quite a lot of possibilities, including the ability to
> build institutional Webs of trust for CAs (ie trust anchors could
> list CAs by reference at one or more levels of indirection), and CAs
> could also describe themselves at their URI.
> 
> Those last two paragraphs are just hints of some possibilities that
> could emerge from moving away from DNs that I can think of. Others
> members of this group will certainly find other possibilities.
> 
> In any case it seems that the time for DNs is passed, and that
> one should perhaps move away from reliance on those and generalise
> this part of TLS.
> 
> Henry
> 
> 
> 
> [1] https://tools.ietf.org/html/draft-thomson-http2-client-certs-01 
> 
> [2] https://www.w3.org/2005/Incubator/webid/spec/identity/#overview 
> 
> [3] https://www.w3.org/2005/Incubator/webid/spec/tls/ 
> 
> 
> ___
> TLS mailing list
> TLS@ietf.org 
> https://www.ietf.org/mailman/listinfo/tls 
> 
> 

___
TLS mailing list
TLS@ietf.org
https://www.ietf.org/mailman/listinfo/tls

Re: [TLS] Generalising DN's to SAN and IAN in TLS1.3?

[Eric Rescorla]

CertificateRequest already allows you to pass an arbitrary number of
extensions by OID.

http://tlswg.github.io/tls13-spec/#certificate-request

What more do you think you need?

-Ekr


On Tue, Mar 8, 2016 at 12:22 AM, Henry Story 
wrote:

> Hi,
>
>
>   I was reading with interest M. Thomson and M. Bishop's
> "Reactive Certificate-Based Client Authentication" draft RFC [1].
> In the section 2.3 "The CERTIFICATE_REQUEST Frame"
>
> [[
>   CA-Count and Certificate-Authorities:  "Certificate-Authorities" is a
>   series of distinguished names of acceptable certificate
>   authorities, represented in DER-encoded [X690] format.  These
>   distinguished names may specify a desired distinguished name for a
>   root CA or for a subordinate CA; thus, this message can be used to
>   describe known roots as well as a desired authorization space.
>   The number of such structures is given by the 16-bit "CA-Count"
>   field, which MAY be zero.  If the "CA-Count" field is zero, then
>   the client MAY send any certificate that meets the rest of the
>   selection criteria in the "CERTIFICATE_REQUEST", unless there is
>   some external arrangement to the contrary.
> ]]
>
> Would it not be possible to extend that so that one could also pass
> issuer Alternative Names, and not just DNs?
>
> Using DNs made sense when it was thought that LDAP and X500 would
> end up being the protocols for global directories. This turned out
> not to be the case. It turned out that DNs were a special case of
> what could be termed a URI (even though DNs are not in URI format).
> And many (most?) URIs can refer to agents, least but not last
> http(s) URLs (See the WebID spec [2] for a nice diagram of how this
> works conceptually and the WebID-TLS spec for one way this can be tied
> to TLS [3]).
>
> If TLS1.3 could start moving away from sole reliance on DNs this
> would open quite a lot of possibilities, including the ability to
> build institutional Webs of trust for CAs (ie trust anchors could
> list CAs by reference at one or more levels of indirection), and CAs
> could also describe themselves at their URI.
>
> Those last two paragraphs are just hints of some possibilities that
> could emerge from moving away from DNs that I can think of. Others
> members of this group will certainly find other possibilities.
>
> In any case it seems that the time for DNs is passed, and that
> one should perhaps move away from reliance on those and generalise
> this part of TLS.
>
> Henry
>
>
>
> [1] https://tools.ietf.org/html/draft-thomson-http2-client-certs-01
> [2] https://www.w3.org/2005/Incubator/webid/spec/identity/#overview
> [3] https://www.w3.org/2005/Incubator/webid/spec/tls/
>
> ___
> TLS mailing list
> TLS@ietf.org
> https://www.ietf.org/mailman/listinfo/tls
>
___
TLS mailing list
TLS@ietf.org
https://www.ietf.org/mailman/listinfo/tls

[TLS] Generalising DN's to SAN and IAN in TLS1.3?

[Henry Story]

Hi,


  I was reading with interest M. Thomson and M. Bishop's 
"Reactive Certificate-Based Client Authentication" draft RFC [1].
In the section 2.3 "The CERTIFICATE_REQUEST Frame" 

[[
  CA-Count and Certificate-Authorities:  "Certificate-Authorities" is a
  series of distinguished names of acceptable certificate
  authorities, represented in DER-encoded [X690] format.  These
  distinguished names may specify a desired distinguished name for a
  root CA or for a subordinate CA; thus, this message can be used to
  describe known roots as well as a desired authorization space.
  The number of such structures is given by the 16-bit "CA-Count"
  field, which MAY be zero.  If the "CA-Count" field is zero, then
  the client MAY send any certificate that meets the rest of the
  selection criteria in the "CERTIFICATE_REQUEST", unless there is
  some external arrangement to the contrary.
]]

Would it not be possible to extend that so that one could also pass
issuer Alternative Names, and not just DNs? 

Using DNs made sense when it was thought that LDAP and X500 would
end up being the protocols for global directories. This turned out
not to be the case. It turned out that DNs were a special case of
what could be termed a URI (even though DNs are not in URI format).
And many (most?) URIs can refer to agents, least but not last
http(s) URLs (See the WebID spec [2] for a nice diagram of how this 
works conceptually and the WebID-TLS spec for one way this can be tied 
to TLS [3]). 

If TLS1.3 could start moving away from sole reliance on DNs this
would open quite a lot of possibilities, including the ability to
build institutional Webs of trust for CAs (ie trust anchors could
list CAs by reference at one or more levels of indirection), and CAs
could also describe themselves at their URI.

Those last two paragraphs are just hints of some possibilities that
could emerge from moving away from DNs that I can think of. Others
members of this group will certainly find other possibilities.

In any case it seems that the time for DNs is passed, and that
one should perhaps move away from reliance on those and generalise
this part of TLS.

Henry



[1] https://tools.ietf.org/html/draft-thomson-http2-client-certs-01
[2] https://www.w3.org/2005/Incubator/webid/spec/identity/#overview
[3] https://www.w3.org/2005/Incubator/webid/spec/tls/

___
TLS mailing list
TLS@ietf.org
https://www.ietf.org/mailman/listinfo/tls