commit 8e684c59d2ba8cd839be2a0bbf21201aec3f910b
parent d50b7a0448378f4f58c1c7fd4c6c3afc45ad872c
Author: Martin Schanzenbach <schanzen@gnunet.org>
Date: Thu, 5 May 2022 16:26:10 +0200
more feedback integrated
Diffstat:
2 files changed, 49 insertions(+), 19 deletions(-)
diff --git a/draft-schanzen-gns.xml b/draft-schanzen-gns.xml
@@ -14,6 +14,7 @@
<!ENTITY RFC5869 PUBLIC '' "http://xml.resource.org/public/rfc/bibxml/reference.RFC.5869.xml">
<!ENTITY RFC5890 PUBLIC '' "http://xml.resource.org/public/rfc/bibxml/reference.RFC.5890.xml">
<!ENTITY RFC5895 PUBLIC '' "http://xml.resource.org/public/rfc/bibxml/reference.RFC.5895.xml">
+<!ENTITY RFC6066 PUBLIC '' "http://xml.resource.org/public/rfc/bibxml/reference.RFC.6066.xml">
<!ENTITY RFC6234 PUBLIC '' "http://xml.resource.org/public/rfc/bibxml/reference.RFC.6234.xml">
<!ENTITY RFC6761 PUBLIC '' "http://xml.resource.org/public/rfc/bibxml/reference.RFC.6761.xml">
<!ENTITY RFC6895 PUBLIC '' "http://xml.resource.org/public/rfc/bibxml/reference.RFC.6895.xml">
@@ -450,8 +451,8 @@
<section anchor="zones" numbered="true" toc="default">
<name>Zones</name>
<t>
- A zone master implementation <bcp14>SHOULD</bcp14> enable the user to
- create and manage zones.
+ A zone master implementation <bcp14>SHOULD</bcp14> enable the zone
+ owners to create and manage zones.
If this functionality is not implemented, names can still be resolved
if zone keys for the initial step in the name resolution are available
(see <xref target="resolution"/>).
@@ -535,6 +536,8 @@
document.
All ztypes <bcp14>MUST</bcp14> be registered as dedicated zone delegation
record types in the GNU Name System Record Types registry (see <xref target="gana"/>).
+ When defining new record types the cryptographic security considerations
+ of this document apply, in particular <xref target="security_cryptography"/>.
</t>
<section anchor="zTLD" numbered="true" toc="default">
<name>Zone Top-Level Domain</name>
@@ -568,7 +571,10 @@
<artwork name="" type="" align="left" alt=""><![CDATA[
zTLD := Base32GNS-Encode(ztype||zkey)
ztype||zkey := Base32GNS-Decode(zTLD)
- ]]></artwork>
+ ]]></artwork>
+ <t>
+ where "||" is the concatenation operator.
+ </t>
<t>
The zTLD can be used as-is as a rightmost label in a GNS name.
If an application wants to ensure DNS compatibility of the name,
@@ -589,7 +595,7 @@ ztype||zkey := Base32GNS-Decode(zTLD)
<!-- FIXME: Is this really really necessary? Really? -->
<artwork name="" type="" align="left" alt=""><![CDATA[
zTLD[126..129].zTLD[63..125].zTLD[0..62]
- ]]></artwork>
+ ]]></artwork>
</section>
<section anchor="revocation" numbered="true" toc="default">
<name>Zone Revocation</name>
@@ -1016,6 +1022,14 @@ zTLD[126..129].zTLD[63..125].zTLD[0..62]
There <bcp14>MAY</bcp14> be inactive records of the same type which have
the SHADOW flag set in order to facilitate smooth key rollovers.
</t>
+ <t>
+ In the following, "||" is the concatenation operator of two byte strings.
+ The algorithm specification uses character strings such as GNS labels or
+ constant values.
+ When used in concatenations or as input to functions the
+ null-terminator of the character strings <bcp14>MUST NOT</bcp14> be
+ included.
+ </t>
<section anchor="gnsrecords_pkey" numbered="true" toc="default">
<name>PKEY</name>
<t>
@@ -1557,9 +1571,11 @@ S-Decrypt(zk,label,expiration,ciphertext):
DNS name of the service to be transmitted over the transport protocol.
In GNS, legacy host name records provide applications the DNS name that
is required to establish a connection to such a service.
- The most common use case is HTTP virtual hosting, where a DNS name must
- be supplied in the HTTP "Host"-header.
- Using a GNS name for the "Host"-header might not work as
+ The most common use case is HTTP virtual hosting and TLS Server Name
+ Indication <xref target="RFC6066"/>, where a DNS name must
+ be supplied in the HTTP "Host"-header and the TLS handshake,
+ respectively.
+ Using a GNS name in those cases might not work as
it might not be globally unique. Furthermore, even if uniqueness is
not an issue, the legacy service might not even be aware of GNS.
</t>
@@ -1688,7 +1704,7 @@ S-Decrypt(zk,label,expiration,ciphertext):
</section>
</section>
<section anchor="publish" numbered="true" toc="default">
- <name>Record Storage</name>
+ <name>Record Encoding</name>
<t>
Any API which allows storing a value under a 512-bit key and retrieving
one or more values from the key can be used by an implementation for record storage.
@@ -2451,6 +2467,12 @@ NICK: john (Supplemental)
<section anchor="security_cryptography" numbered="true" toc="default">
<name>Cryptography</name>
<t>
+ The following considerations provide background on the design choices
+ of the ztypes specified in this document.
+ When specifying new ztypes as per <xref target="zones"/>, the same
+ considerations apply.
+ </t>
+ <t>
GNS PKEY zone keys use ECDSA over Ed25519.
This is an unconventional choice,
as ECDSA is usually used with other curves. However, standardized
@@ -2934,7 +2956,7 @@ Purpose | Name | References | Comment
<references>
<name>Informative References</name>
&RFC4033;
- <!-- &RFC6781; -->
+ &RFC6066;
&RFC7363;
&RFC8324;
&RFC8806;
diff --git a/rs_review_202204.txt b/rs_review_202204.txt
@@ -33,8 +33,7 @@ Each definition might be more clear if the repetition of the term were omitted a
In the "Name" definition, do you mean "applications MAY *require* that ... "?
-===== I think we could also formulate it like that. But that would be a normative statement for the application
-as opposed to a normative statement what the GNS impl. should expect.
+===== I think we could also formulate it like that. But that would be a normative statement for the application as opposed to a normative statement what the GNS impl. should expect. => No change.
Using "Zone type" as the term for cipher and encoding seemed strange to me, why not "Zone Representation" or something shorter? One or two examples before the terms, with text identifying the parts in the definition, would be helpful.
@@ -46,7 +45,7 @@ Sec 3, "cryptographically secured zones" is redundant, aren't all zones cryptogr
Last sentence of first paragraph repeats the definition we just read.
-===== I don't think so.
+===== Unable to find this issue in the text itself.
Second para "A zone can be populated by its owner with ..." A clarification that the distributed storage isn't required by the protocol would be useful.
@@ -54,21 +53,23 @@ Second para "A zone can be populated by its owner with ..." A clarification that
The text introducing figure 2 makes me think it's going to show all the flows for a recursive resolution, and it is rather a high-level view.
-===== That is what the figure description sais and I think this is enough for an "Overview".
+===== That is what the figure description sais and I think this is enough for an "Overview".
Sec 4, which user creates and manages zones? The Zone admin? The end-user?
-==== TODO
+==== Zone owners. Clarified.
The forward ref to Section 5.1 should be before the list of algorithms in my opinion.
-==== TODO. After first review I disagree as it refers to the functions which are introduced above.
+==== After review I disagree as it refers to the functions which are introduced above that text.
Sec 4.1, maybe say "||" is the concatenation operator and that "[a..b]" is the semi-closed range from A to B-1
+==== Clarified the use of the operators.
+
Sec 4.2, "strictly monotonically increasing order" I think monotonically is wrong.
-====== I am 95% sure that monotonically is correct when we talk about a function. Not sure about order.
+====== I am 99% sure that monotonically is correct and the important part.
Sec 5, nice to block of IANA numbers for interop. Are timestamps in GNS always 64bit microsec since 1/1/1970? Maybe add a definition or something at the top?
@@ -82,18 +83,23 @@ Sec 5.1.1, the crypto heart starts to appear. ;) This all looks okay to me, and
Sec 5.1.2, "highest 32 bytes" maybe "top-most"? Nit. SignDerived is clever. Consider asking CFRG to review the crypto in Sec 5.
+======= We could do that, but then again we had a review by djb already.
+
Sec 5.2.1+5.2.2, have you thought of allowing multiple REDIRECT records to achieve some kind of load-balancing or other?
-======= Q: Maybe we want this? CNAME did not allow this, but it seems to be used in practice occasionally https://docstore.mik.ua/orelly/networking_2ndEd/dns/ch10_07.htm#dns4-CHP-10-SECT-7.1
+======= This is a CNAME-derivative. It seems that CNAMEs do not support this either: https://docstore.mik.ua/orelly/networking_2ndEd/dns/ch10_07.htm#dns4-CHP-10-SECT-7.1
Sec 5.3.1, I would think TLS SNI value is also common for LEHO data.
+===== Added ref to RFC6066 SNI
+
Sec 5.3.3, not unlike the new SVCB RR type?
+==== Yeah, bit is still draft status (?).
+
Sec 6, is this whole storage strictly necessary for interop? You could split it off into a separate document. There's not enough here to implement the storage. What happens to resolution if the GET() fails? I assume that's discussed.
-======= FIXME: We may want to consider renaming Section 6 and/or move it into a subsection of Section 5 as we do not really define the storage here.
-======= Record Wire Format / Record Encoding
+======= Renamed section to "Record Encoding" in order to avoid storage confusion.
Sec 7, the application filters record sets. Oh, that is *VERY* interesting. And that picture is starting to look familiar. Third time?
@@ -101,6 +107,8 @@ Sec 7.1+7.2+7.3 very nice, very clear, I could implement this (modulo the storag
Sec 9.3 seems very important. Somewhere up near the front you should forward-link to it. And in the security considerations, backlink.
+===== Added.
+
Sec 9.4, it's difficult because it requires law enforcement, etc., to get the private key? Is that really so hard? https://xkcd.com/538/
==== No, because of the user-centric design for delegations. TODO: Clarify again in text?
