> -----Original Message-----
> From: dns-privacy [mailto:[email protected]] On Behalf Of
> Alexander Mayrhofer
> Sent: Monday, July 25, 2016 5:04 PM
> To: [email protected]
> Subject: Re: [dns-privacy] I-D Action: draft-ietf-dprive-dnsodtls-07.txt
> 
> 
> Hi,
> 
> i've reviewed the DNSoD document in version -07, and i think that the
> document is in good shape to proceed to WGLC (I suppose we'd get one or
> more DISCUSSes from the SEC AD during the IESG review, though ;). I do only
> have a couple of minor comments:
> 
> [Disclaimer: I don't know much about DTLS, admitted]
> 
> 1) The introduction explicitly lists use cases in which DNSoD provides
> confidential DNS communication (eg. stub to recursive). What is the reason
> for "limiting" this to these use cases - and why would DNSoD not provide this
> property in resolver-authoritative communications? I know the charter does
> currently not address that side of the DNS,  but I don't see any reason why it
> wouldn't work there. Proposal: change text so that is doesn't seem to exclude
> the other use cases.

Discussion of encrypting DNS messages in other use cases has just started in 
the WG and requires re-charter. 
https://tools.ietf.org/html/rfc7858#section-1 in the last paragraph explains 
that other use cases are out of scope of the doc.

> 
> 2) I think a sentence that the (wire) format of the DNS messages remains
> unchanged would be helpful.

Added.

> 
> 3) Editorial: In some cases, adding the document title in front of a reference
> would improve readability. Eg. "...which can be done with TLS Session
> Resumption without server state (RFC5077)." Instead of  "...which can be done
> with [RFC5077]."

Fixed.

> 
> 4) Do I understand correctly that in section 4 we explicitely rule out
> fragmentation, by limiting response sizes to the Path MTU (rather than the
> UDP payload size indicated via EDNS0 by the client)? 

Yes, RFC6347 in section 4.1.1 discusses that each DTLS record MUST fit within a 
single datagram.

> If yes so, why does this
> only apply to the server ?

Because only the DNS response can be larger than 512 bytes, see RFC6891.  

> Theoretically, the client could also create a query
> that exceeds the path MTU... Furthermore, (again, little knowledge of DTLS) is
> ruling out fragmentation a limitation of using DTLS, or was it a design 
> choice? 

DTLS handles fragmentation and reassembly only for handshake messages. In 
previous versions of this draft we had added support for fragmentation, 
but removed it based on discussion in the WG.

> I
> think, generally, this section is crucial and should be restructured, eg. 
> splitting
> off the size calculations from the server requirements.
> 
> 5) The statement "The largest gain for privacy is to protect the communication
> between the DNS client (the end user's machine) and its caching resolver."
> Comes a bit out of the blue - a reference here would be good. Or,
> alternatively, move this to the introduction or remove it entirely?

Moved the above line to Introduction.

> 
> 6) Section 6: Replace "plain DNS" with "plain text DNS"? Furthermore, change
> "as modern networks require DNS" to "as modern applications typically
> require DNS"

Done.

> 
> 7) Some minor restructuring of the Security Considerations section might do
> good. Eg. splitting the first sentence, like "with a ciphersuite offering
> confidentiality protection. Guidance given..." would make it clearer.

Updated.

Thanks,
-Tiru

> 
> Best,
> Alex
> 
> 
> 
> 
> _______________________________________________
> dns-privacy mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/dns-privacy

_______________________________________________
dns-privacy mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dns-privacy

Reply via email to