Update to DNAME Redirection in the DNS

DNAME is a DNS Resource Record type originally defined in RFC 2672 . DNAME provides redirection from a part of the DNS name tree to another part of the DNS name tree. The DNAME RR and the CNAME RR cause a lookup to (potentially) return data corresponding to a domain name different from the queried domain name. The difference between the two resource records is that the CNAME RR directs the lookup of data at its owner to another single name, a DNAME RR directs lookups for data at descendents of its owner's name to corresponding names under a different (single) node of the tree. Take for example, looking through a zone (see RFC 1034, section 4.3.2, step 3) for the domain name "foo.example.com" and a DNAME resource record is found at "example.com" indicating that all queries under "example.com" be directed to "example.net". The lookup process will return to step 1 with the new query name of "foo.example.net". Had the query name been "www.foo.example.com" the new query name would be "www.foo.example.net". This document is a revision of the original specification of DNAME in RFC 2672. DNAME was conceived to help with the problem of maintaining address-to-name mappings in a context of network renumbering. With a careful set-up, a renumbering event in the network causes no change to the authoritative server that has the address-to-name mappings. Examples in practice are classless reverse address space delegations. Another usage of DNAME lies in aliasing of name spaces. For example, a zone administrator may want sub-trees of the DNS to contain the same information. Examples include punycode alternates for domain spaces. This revision to DNAME does not change the wire format or the handling of DNAME Resource Records by existing software. A new UD (Understand DNAME) bit in the EDNS flags field can be used to signal that CNAME synthesis is not needed. Discussion is added on problems that may be encountered when using DNAME.

The DNAME RR has mnemonic DNAME and type code 39 (decimal). It is not class-sensitive.

Its RDATA is comprised of a single field, <target>, which contains a fully qualified domain name that must be sent in uncompressed form , . The <target> field MUST be present. The presentation format of <target> is that of a domain name . <owner> <ttl> <class> DNAME <target> The effect of the DNAME RR is the substitution of the record's <target> for its owner name, as a suffix of a domain name. This substitution has to be applied for every DNAME RR found in the resolution process, which allows fairly lengthy valid chains of DNAME RRs. Details of the substitution process, methods to avoid conflicting resource records, and rules for specific corner cases are given in the following subsections.

When following RFC 1034 , section 4.3.2's algorithm's third step, "start matching down, label by label, in the zone" and a node is found to own a DNAME resource record a DNAME substitution occurs. The name being sought may be the original query name or a name that is the result of a CNAME resource record being followed or a previously encountered DNAME. As is the case of finding a CNAME resource record or NS resource record set, the processing of a DNAME will happen prior to finding the desired domain name. A DNAME substitution is performed by replacing the suffix labels of the name being sought matching the owner name of the DNAME resource record with the string of labels in the RDATA field. The matching labels end with the root label in all cases. Only whole labels are replaced. See the table of examples for common cases and corner cases.

In the table below, the QNAME refers to the query name. The owner is the DNAME owner domain name, and the target refers to the target of the DNAME record. The result is the resulting name after performing the DNAME substitution on the query name. "no match" means that the query did not match the DNAME and thus no substitution is performed and a possible error message is returned (if no other result is possible). In the examples below, 'cyc' and 'shortloop' contain loops. QNAME owner DNAME target result ---------------- -------------- -------------- ----------------- com. example.com. example.net. <no match> example.com. example.com. example.net. <no match> a.example.com. example.com. example.net. a.example.net. a.b.example.com. example.com. example.net. a.b.example.net. ab.example.com. b.example.com. example.net. <no match> foo.example.com. example.com. example.net. foo.example.net. a.x.example.com. x.example.com. example.net. a.example.net. a.example.com. example.com. y.example.net. a.y.example.net. cyc.example.com. example.com. example.com. cyc.example.com. cyc.example.com. example.com. c.example.com. cyc.c.example.com. shortloop.x.x. x. . shortloop.x. shortloop.x. x. . shortloop. It is possible for DNAMEs to form loops, just as CNAMEs can form loops. DNAMEs and CNAMEs can chain together to form loops. A single corner case DNAME can form a loop. Resolvers and servers should be cautious in devoting resources to a query, but be aware that fairly long chains of DNAMEs may be valid. Zone content administrators should take care to insure that there are no loops that could occur when using DNAME or DNAME/CNAME redirection. The domain name can get too long during substitution. For example, suppose the target name of the DNAME RR is 250 octets in length (multiple labels), if an incoming QNAME that has a first label over 5 octets in length, the result of the result would be a name over 255 octets. If this occurs the server returns an RCODE of YXDOMAIN . The DNAME record and its signature (if the zone is signed) are included in the answer as proof for the YXDOMAIN (value 6) RCODE.

Unlike a CNAME RR, a DNAME RR redirects DNS names subordinate to its owner name; the owner name of a DNAME is not redirected itself. The domain name that owns a DNAME record is allowed to have other resource record types at that domain name, except DNAMEs or CNAMEs. This means that DNAME RRs are not allowed at the parent side of a delegation point but are allowed at a zone apex. The reason for this decision was that one can have a DNAME at the zone apex. There still is a need to have the customary SOA and NS resource records at the zone apex. This means that DNAME does not mirror a zone completely, as it does not mirror the zone apex. These rules also allow DNAME records to be queried through RFC 1034 compliant, DNAME-unaware caches.

Resource records MUST NOT exist at any domain name subordinate to the owner of a DNAME RR. To get the contents for names subordinate to that owner, the DNAME redirection must be invoked and the resulting target queried. A server MAY refuse to load a zone that has data at a domain name subordinate to a domain name owning a DNAME RR. If the server does load the zone, those names below the DNAME RR will be occluded, RFC 2136 , section 7.18. Also a server SHOULD refuse to load a zone subordinate to the owner of a DNAME record in the ancestor zone. See for further discussion related to dynamic update. DNAME is a singleton type, meaning only one DNAME is allowed per name. The owner name of a DNAME can only have one DNAME RR, and no CNAME RRs can exist at that name. These rules make sure that for a single domain name only one redirection exists, and thus no confusion which one to follow. A server SHOULD refuse to load a zone that violates these rules.

The DNAME owner name can be compressed like any other owner name. The DNAME RDATA target name MUST NOT be sent out in compressed form, so that a DNAME RR can be treated as an unknown type . Although the previous DNAME specification (that is obsoleted by this specification) talked about signaling to allow compression of the target name, such signaling has never been specified and this document also does not specify this signaling behavior. RFC 2672 (obsoleted by this document) stated that the EDNS version had a meaning for understanding of DNAME and DNAME target name compression. This document revises RFC 2672, in that there is no EDNS version signaling for DNAME. However, the flags section of EDNS(0) is updated with a Understand-DNAME flag by this document (See Section 3.3).

The DNAME RR causes type NS additional section processing.

When preparing an response, a server upon performing a DNAME substitution will in all cases include the DNAME RR used in the answer section. A CNAME RR record with TTL equal to the corresponding DNAME RR is synthesized and included in the answer section for old resolvers. The owner name of the CNAME is the QNAME of the query. DNSSEC , , says that the synthesized CNAME does not have to be signed. The DNAME has an RRSIG and a validating resolver can check the CNAME against the DNAME record and validate the DNAME record. Resolvers MUST be able to handle a synthesized CNAME TTL of zero or equal to the TTL of the corresponding DNAME record. A TTL of zero means that the CNAME can be discarded immediately after processing the answer. DNAME aware resolvers can set the Understand-DNAME (UD bit) to receive a response with only the DNAME RR and no synthesized CNAMEs. The UD bit is part of the EDNS extended RCODE and Flags field. It is used to omit server processing, transmission and resolver processing of unsigned synthesized CNAMEs. Resolvers can set this in a query to request omission of the synthesized CNAMEs. Servers copy the UD bit to the response, and can omit synthesized CNAMEs from the answer. Older resolvers do not set the UD bit, and older servers do not copy the UD bit to the answer, and will not omit synthesized CNAMEs.

Updated EDNS extended RCODE and Flags field. +0 (MSB) +1 (LSB) +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ 0: | EXTENDED-RCODE | VERSION | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ 2: |DO|UD| Z | +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+ Servers MUST be able to answer a query for a synthesized CNAME. Like other query types this invokes the DNAME, and synthesizes the CNAME into the answer.

Below the server algorithm, which appeared in RFC 2672 Section 4.1, is expanded to handle the UD (Understand-DNAME) bit. Set or clear the value of recursion available in the response depending on whether the name server is willing to provide recursive service. If recursive service is available and requested via the RD bit in the query, go to step 5, otherwise step 2. Search the available zones for the zone which is the nearest ancestor to QNAME. If such a zone is found, go to step 3, otherwise step 4. Start matching down, label by label, in the zone. The matching process can terminate several ways: If the whole of QNAME is matched, we have found the node. If the data at the node is a CNAME, and QTYPE does not match CNAME, copy the CNAME RR into the answer section of the response, change QNAME to the canonical name in the CNAME RR, and go back to step 1. Otherwise, copy all RRs which match QTYPE into the answer section and go to step 6. If a match would take us out of the authoritative data, we have a referral. This happens when we encounter a node with NS RRs marking cuts along the bottom of a zone. Copy the NS RRs for the sub-zone into the authority section of the reply. Put whatever addresses are available into the additional section, using glue RRs if the addresses are not available from authoritative data or the cache. Go to step 4. If at some label, a match is impossible (i.e., the corresponding label does not exist), look to see whether the last label matched has a DNAME record. If a DNAME record exists at that point, copy that record into the answer section. If substitution of its <target> for its <owner> in QNAME would overflow the legal size for a <domain- name>, set RCODE to YXDOMAIN and exit; otherwise perform the substitution and continue. If the EDNS OPT record is present in the query and the UD bit is set, the server MAY copy the UD bit to the answer EDNS OPT record, and omit CNAME synthesis. Else the server MUST synthesize a CNAME record as described above and include it in the answer section. Go back to step 1. If there was no DNAME record, look to see if the "*" label exists. If the "*" label does not exist, check whether the name we are looking for is the original QNAME in the query or a name we have followed due to a CNAME or DNAME. If the name is original, set an authoritative name error in the response and exit. Otherwise just exit. If the "*" label does exist, match RRs at that node against QTYPE. If any match, copy them into the answer section, but set the owner of the RR to be QNAME, and not the node with the "*" label. If the data at the node with the "*" label is a CNAME, and QTYPE doesn't match CNAME, copy the CNAME RR into the answer section of the response changing the owner name to the QNAME, change QNAME to the canonical name in the CNAME RR, and go back to step 1. Otherwise, Go to step 6. Start matching down in the cache. If QNAME is found in the cache, copy all RRs attached to it that match QTYPE into the answer section. If QNAME is not found in the cache but a DNAME record is present at an ancestor of QNAME, copy that DNAME record into the answer section. If there was no delegation from authoritative data, look for the best one from the cache, and put it in the authority section. Go to step 6. Use the local resolver or a copy of its algorithm to answer the query. Store the results, including any intermediate CNAMEs and DNAMEs, in the answer section of the response. Using local data only, attempt to add other RRs which may be useful to the additional section of the query. Exit. Note that there will be at most one ancestor with a DNAME as described in step 4 unless some zone's data is in violation of the no-descendants limitation in section 3. An implementation might take advantage of this limitation by stopping the search of step 3c or step 4 when a DNAME record is encountered.

The use of DNAME in conjunction with wildcards is discouraged . Thus records of the form "*.example.com DNAME example.net" SHOULD NOT be used. The interaction between the expansion of the wildcard and the redirection of the DNAME is non-deterministic. Because the processing is non-deterministic, DNSSEC validating resolvers may not be able to validate a wildcarded DNAME. A server MAY give a warning that the behavior is unspecified if such a wildcarded DNAME is loaded. The server MAY refuse it, refuse to load or refuse dynamic update.

Recursive caching name servers can encounter data at names below the owner name of a DNAME RR, due to a change at the authoritative server where data from before and after the change resides in the cache. This conflict situation is a transitional phase, that ends when the old data times out. The caching name server can opt to store both old and new data and treat each as if the other did not exist, or drop the old data, or drop the longer domain name. In any approach, consistency returns after the older data TTL times out. Recursive caching name servers MUST perform CNAME synthesis on behalf of DNAME-ignorant clients. A recursive caching name server that understands DNAMEs can send out queries on behalf of clients with the UD bit set (See ). After receiving the answers the recursive caching name server sends replies to DNAME ignorant clients that include DNAMEs and synthesized CNAMEs.

In , in Section 10.3., the discussion on MX and NS records touches on redirection by CNAMEs, but this also holds for DNAMEs.

Excerpt from 10.3. MX and NS records (in RFC 2181). The domain name used as the value of a NS resource record, or part of the value of a MX resource record must not be an alias. Not only is the specification clear on this point, but using an alias in either of these positions neither works as well as might be hoped, nor well fulfills the ambition that may have led to this approach. This domain name must have as its value one or more address records. Currently those will be A records, however in the future other record types giving addressing information may be acceptable. It can also have other RRs, but never a CNAME RR. The DNAME RR is discussed in RFC 3363, section 4, on A6 and DNAME. The opening premise of this section is demonstrably wrong, and so the conclusion based on that premise is wrong. In particular, deprecates the use of DNAME in the IPv6 reverse tree, which is then carried forward as a recommendation in . Based on the experience gained in the meantime, should be revised, dropping all constraints on having DNAME RRs in these zones. This would greatly improve the manageability of the IPv6 reverse tree. These changes are made explicit below.

In , the paragraph "The issues for DNAME in the reverse mapping tree appears to be closely tied to the need to use fragmented A6 in the main tree: if one is necessary, so is the other, and if one isn't necessary, the other isn't either. Therefore, in moving RFC 2874 to experimental, the intent of this document is that use of DNAME RRs in the reverse tree be deprecated." is to be replaced with the word "DELETED".

In , the reference to DNAME was left in as an editorial oversight. The paragraph "Those nodes are NOT RECOMMENDED to support the experimental A6 and DNAME Resource Records [RFC3363]."

is to be replaced by "Those nodes are NOT RECOMMENDED to support the experimental A6 Resource Record [RFC3363]."

There are several issues to be aware of about the use of DNAME.

The names listed as target names of MX, NS, PTR and SRV records must be canonical hostnames. This means no CNAME or DNAME redirection may be present during DNS lookup of the address records for the host. This is discussed in RFC 2181 , section 10.3, and RFC 1912 , section 2.4. For SRV see RFC 2782 page 4. The upshot of this is that although the lookup of a PTR record can involve DNAMEs, the name listed in the PTR record can not fall under a DNAME. The same holds for NS, SRV and MX records. For example, when punycode alternates for a zone use DNAME then the NS, MX, SRV and PTR records that point to that zone must use names without punycode in their RDATA. What must be done then is to have the domain names with DNAME substitution already applied to it as the MX, NS, PTR, SRV data. These are valid canonical hostnames.

DNAME records can be added, changed and removed in a zone using dynamic update transactions. Adding a DNAME RR to a zone occludes any domain names that may exist under the added DNAME. A server MUST reject a dynamic update message that attempts to add a DNAME RR at a name that already has a CNAME RR or another DNAME RR associated with that name.

When a validator checks the NSEC RRs returned on a name error response, it SHOULD check that the DNAME bit is not set. If the DNAME bit is set then the DNAME substitution should have been done, but has not.

Below are examples of why DNSSEC validators MUST understand DNAME.

;; Header: QR AA DO RCODE=3(NXDOMAIN) ;; Question foo.bar.example.com. IN A ;; Answer bar.example.com. NSEC dub.example.com. A DNAME bar.example.com. RRSIG NSEC [valid signature] If this is the response, then only by understanding that the DNAME bit means that foo.bar.example.com needed to have been redirected by the DNAME, the validator can see that it is a BOGUS reply from an attacker that collated existing records from the DNS to create a confusing reply. If the DNAME bit had not been set in the NSEC record above then the answer would have validated as a correct name error response.

;; Header: QR AA DO RCODE=3(NXDOMAIN) ;; Question cee.example.com. IN A ;; Answer bar.example.com. NSEC dub.example.com. A DNAME bar.example.com. RRSIG NSEC [valid signature] This reply has the same NSEC records as the example above, but with this query name (cee.example.com), the answer is validated, because 'cee' does not get redirected by the DNAME at 'bar'.

;; Header: QR AA DO RCODE=0(NOERROR) ;; Question foo.bar.example.com. IN A ;; Answer bar.example.com. DNAME bar.example.net. bar.example.com. RRSIG DNAME [valid signature] foo.bar.example.com. CNAME foo.bar.example.net. The answer shown above has the synthesized CNAME included. However, the CNAME has no signature, since the server does not sign online. So it cannot be trusted. It could be altered by an attacker to be foo.bar.example.com CNAME bla.bla.example. The DNAME record does have its signature included, since it does not change for every query name. The validator must verify the DNAME signature and then recursively resolve further to query for the foo.bar.example.net A record.

The DNAME Resource Record type code 39 (decimal) originally has been registered by [RFC2672]. IANA should update the DNS resource record registry to point to this document for RR type 39. This draft requests the second highest bit in the EDNS flags field for the Understand-DNAME (UD) flag as described in Section 3.1.

DNAME redirects queries elsewhere, which may impact security based on policy and the security status of the zone with the DNAME and the redirection zone's security status. If a validating resolver accepts wildcarded DNAMEs, this creates security issues. Since the processing of a wildcarded DNAME is non-deterministic and the CNAME that was substituted by the server has no signature, the resolver may choose a different result than what the server meant, and consequently end up at the wrong destination. Use of wildcarded DNAMEs is discouraged in any case . A validating resolver MUST understand DNAME, according to . In examples are given that illustrate this need.

The authors of this draft would like to acknowledge Matt Larson for beginning this effort to address the issues related to the DNAME RR type. The authors would also like to acknowledge Paul Vixie, Ed Lewis, Mark Andrews, Mike StJohns, Niall O'Reilly, Sam Weiler, Alfred Hoenes and Kevin Darcy for their review and comments on this document.