Re: [regext] Alexey Melnikov's Discuss on draft-ietf-regext-login-security-07: (with DISCUSS and COMMENT)

["Alexey Melnikov" <[e-mail]>]

Hi James,

(Top post)

Your changes look good to me.

Best Regards,
Alexey

On Mon, Jan 27, 2020, at 4:40 PM, Gould, James wrote:
> Alexey,

> 

> My responses are embedded below.

> 

> --

> 

> JG

> 

> 

> 

> James Gould

> Distinguished Engineer

> [e-mail] <applewebdata://13890C55-AAE8-4BF3-A6CE-B4BA42740803/[e-mail]>

> 

> 703-948-3271

> 12061 Bluemont Way

> Reston, VA 20190

> 

> Verisign.com <http://verisigninc.com/>

> 

>  On 1/27/20, 9:02 AM, "Alexey Melnikov" <[e-mail]> wrote:

> 

>  Hi James,

> 

>  On Thu, Jan 23, 2020, at 9:29 PM, Gould, James wrote:

>  > Alexey,

> 

>  [snip]

> 

>  > 

>  > Thank you for this document. I have several small comments similar

>  > to what was

>  > raised by Roman and Ben:

>  > 

>  > 1) In 4.1:

>  > 

>  > <loginSec:userAgent>: OPTIONAL client user agent that identifies the

>  > client application software, technology, and operating system

>  > used by the server to identify functional or security

>  > constraints, current security issues, and potential future

>  > functional or security issues for the client. The

>  > <loginSec:userAgent> element MUST contain at least one of the

>  > following child elements:

>  > 

>  > <loginSec:app>: OPTIONAL name of the client application software

>  > with version if available, such as the name of the client SDK

>  > "EPP SDK 1.0.0".

>  > <loginSec:tech>: OPTIONAL technology used for the client

>  > software with version if available, such as "Java 11.0.2".

>  > <loginSec:os>: OPTIONAL client operating system used with

>  > version if available, such as "x86_64 Mac OS X 10.11.6".

>  > 

>  > Is there a registry of allowed values or at least some instructions how to

>  > construct these values? There are probably several existing IETF registries

>  > that can be reused.

>  >

>  > JG - I'm not aware of any registries that exist that describes how to

>  > construct these values. I believe to address Roman's, Ben's and your

>  > feedback, I will provide an example of how to construct these values.

> 

>  I am not insisting that you should use these, but some examples:

> 

>  https://www.iana.org/assignments/operating-system-names/operating-system-names.xhtml#operating-system-names-1

> 

>  and probably more interesting:

> 

>  https://www.iana.org/assignments/iodef2/iodef2.xhtml#softwarereference-dtype

> 

>  The latter points to NIST's Common Platform Enumeration and ISO 19770 software identification (SWID).

> 

> JG - I believe inclusion of the example ABNF for constructing the <loginSec:userAgent> child element values will provide the needed clarification. It looks like you feel the same way based on your response to Roman's thread. I'll address the usage of the information.

> 

> 

>  > 

>  > If these values are not supposed to be used by servers for anything

>  > other than

>  > logging (i.e. if they can't be used to work around bugs), then the

>  > document

>  > needs to say that.

>  >

>  > JG - The servers can leverage this information for more than logging;

>  > although logging is the most common use case. The most useful element

>  > for identification is the <loginSec:app>, where if there is a known

>  > client application such as an EPP SDK, the server can key off of the

>  > EPP SDK version to proactively identify potential security issues to

>  > report back to the client. The server may already know the client

>  > application patterns or can identify the client application patterns

>  > using the logs to create rules for specific clients applications. The

>  > additional <loginSec:tech> and <loginSec:os> elements are useful to

>  > identify future security policy issues, such as deprecating or removing

>  > TLS cipher suites or TLS protocols. The short answer to your question

>  > is that the server can utilize the elements for logging and for driving

>  > security event rules, but this is beyond the scope of the extension

>  > specification. 

> 

>  I think saying something along these lines would be helpful.

> 

> JG - To address this, I can update the description of the <loginSec:userAgent> as:

> 

> OPTIONAL client user agent information that identifies the client application software, technology, and operating system used by the server to identify functional or security constraints, current security issues, and potential future functional or security issues for the client. *The server may use the information for real-time identification and client notification of security issues, such as keying off of the client application software for executing security rule checks. The server may capture the information to identify future security policy issues, such as deprecating or removing TLS cipher suites or TLS protocols.* The <loginSec:userAgent> element MUST contain at least one of the following child elements:

> 

> 

>  Best Regards,

>  Alexey

>