Internet Engineering Task Force (IETF) C. Contavalli Request for Comments: 7871 W. van der Gaast Category: Informational Google ISSN: 2070-1721 D. Lawrence Akamai Technologies W. Kumari Google May 2016
Client Subnet in DNS Queries
Abstract
This document describes an Extension Mechanisms for DNS (EDNS0) option that is in active use to carry information about the network that originated a DNS query and the network for which the subsequent response can be cached. Since it has some known operational and privacy shortcomings, a revision will be worked through the IETF for improvement.
Status of This Memo
This document is not an Internet Standards Track specification; it is published for informational purposes.
This document is a product of the Internet Engineering Task Force (IETF). It represents the consensus of the IETF community. It has received public review and has been approved for publication by the Internet Engineering Steering Group (IESG). Not all documents approved by the IESG are a candidate for any level of Internet Standard; see Section 2 of RFC 5741.
Information about the current status of this document, any errata, and how to provide feedback on it may be obtained at http://www.rfc-editor.org/info/rfc7871.
Contavalli, et al. Informational [Page 1]
RFC 7871 Client Subnet in DNS Queries May 2016
Copyright Notice
Copyright (c) 2016 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
Many Authoritative Nameservers today return different responses based on the perceived topological location of the user. These servers use the IP address of the incoming query to identify that location.
Since most queries come from Intermediate Recursive Resolvers, the source address is that of the Recursive Resolver rather than of the query originator.
Traditionally, and probably still in the majority of instances, Recursive Resolvers are reasonably close in the topological sense to the Stub Resolvers or Forwarding Resolvers that are the source of queries. For these resolvers, using their own IP address is sufficient for Authoritative Nameservers that tailor responses based upon location of the querier.
Increasingly, though, a class of Recursive Resolvers has arisen that handles query sources that are often not topologically close. The motivation for having such Centralized Resolvers varies but is usually because of some enhanced experience, such as greater cache security or applying policies regarding where users may connect. (Although political censorship usually comes to mind here, the same actions may be used by a parent when setting controls on where a minor may connect.) Similarly, many ISPs and other organizations use a Centralized Resolver infrastructure that can be distant from the clients the resolvers serve. These cases all lead to less than desirable responses from topology-sensitive Authoritative Nameservers.
This document defines an EDNS0 [RFC6891] option to convey network information that is relevant to the DNS message. It will carry sufficient network information about the originator for the Authoritative Nameserver to tailor responses. It will also provide for the Authoritative Nameserver to indicate the scope of network addresses for which the tailored answer is intended. This EDNS0 option is intended for those Recursive Resolvers and Authoritative Nameservers that would benefit from the extension and not for general purpose deployment. This is completely optional and can safely be ignored by servers that choose not to implement or enable it.
This document also includes guidelines on how best to cache those results, and it provides recommendations on when this protocol extension should be used.
At least a dozen different client and server implementations have been written based on earlier draft versions of this specification. The protocol is in active production use today. While the
Contavalli, et al. Informational [Page 4]
RFC 7871 Client Subnet in DNS Queries May 2016
implementations interoperate, there is varying behavior around edge cases that were poorly specified. Known incompatibilities are described in this document, and the authors believe that it is better to describe the system as it is working today, even if not everyone agrees with the details of the original specification ([VANDERGAAST]). The alternative is an undocumented and proprietary system.
A revised proposal to improve upon the minor flaws in this protocol will be forthcoming to the IETF.
If we were just beginning to design this mechanism, and not documenting existing protocol, it is unlikely that we would have done things exactly this way.
The IETF is actively working on enhancing DNS privacy [DPRIVE_Working_Group] and the reinjection of metadata [METADATA] has been identified as a problematic design pattern.
As noted above however, this document primarily describes existing behavior of a deployed method to further the understanding of the Internet community.
We recommend that the feature be turned off by default in all nameserver software, and that operators only enable it explicitly in those circumstances where it provides a clear benefit for their clients. We also encourage the deployment of means to allow users to make use of the opt-out provided. Finally, we recommend that others avoid techniques that may introduce additional metadata in future work, as it may damage user trust.
Regrettably, support for the opt-out provisions of this specification are currently limited. Only one stub resolver, getdns, is known to be able to originate queries with anonymity requested, and as yet no applications are known to be able to indicate that user preference to the stub resolver.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119].
Client: A Stub Resolver, Forwarding Resolver, or Recursive Resolver. A client to a Recursive Resolver or a Forwarding Resolver.
Server: A Forwarding Resolver, Recursive Resolver, or Authoritative Nameserver.
Stub Resolver: A simple DNS protocol implementation on the client side as described in [RFC1034], Section 5.3.1. A client to a Recursive Resolver or a Forwarding Resolver.
Authoritative Nameserver: A nameserver that has authority over one or more DNS zones. These are normally not contacted by Stub Resolver or end user clients directly but by Recursive Resolvers. Described in [RFC1035], Section 6.
Recursive Resolver: A nameserver that is responsible for resolving domain names for clients by following the domain's delegation chain. Recursive Resolvers frequently use caches to be able to respond to client queries quickly. Described in [RFC1035], Section 7.
Forwarding Resolver: A nameserver that does not do iterative resolution itself, but instead passes that responsibility to another Recursive Resolver, called a "Forwarder" in [RFC2308], Section 1.
Intermediate Nameserver: Any nameserver in between the Stub Resolver and the Authoritative Nameserver, such as a Recursive Resolver or a Forwarding Resolver.
Centralized Resolvers: Intermediate Nameservers that serve a topologically diverse network address space.
Tailored Response: A response from a nameserver that is customized for the node that sent the query, often based on performance (i.e., lowest latency, least number of hops, topological distance, etc.).
Topologically Close: Refers to two hosts being close in terms of the number of hops or the time it takes for a packet to travel from one host to the other. The concept of topological distance is only loosely related to the concept of geographical distance: two
Contavalli, et al. Informational [Page 6]
RFC 7871 Client Subnet in DNS Queries May 2016
geographically close hosts can still be very distant from a topological perspective, and two geographically distant hosts can be quite close on the network.
For a more comprehensive treatment of DNS terms, please see [RFC7719].
The general idea of this document is to provide an EDNS0 option to allow Recursive Resolvers, if they are willing, to forward details about the origin network from which a query is coming when talking to other nameservers.
The format of the edns-client-subnet (ECS) EDNS0 option is described in Section 6 and is meant to be added in queries sent by Intermediate Nameservers in a way that is transparent to Stub Resolvers and end users, as described in Section 7.1. ECS is only defined for the Internet (IN) DNS class.
As described in Section 7.2, an Authoritative Nameserver could use ECS as a hint to the end user's network location and provide a better answer. Its response would also contain an ECS option, clearly indicating that the server made use of this information, and that the answer is tied to the client's network.
As described in Section 7.3, Intermediate Nameservers would use this information to cache the response.
Some Intermediate Nameservers may also have to be able to forward ECS queries they receive, as described in Section 7.5.
The mechanisms provided by ECS raise various security-related concerns related to cache growth, the ability to spoof EDNS0 options, and privacy. Section 11 explores various mitigation techniques.
The expectation, however, is that this option will primarily be used between Recursive Resolvers and Authoritative Nameservers that are sensitive to network location issues. Most Recursive Resolvers, Authoritative Nameservers, and Stub Resolvers will never need to know about this option and will continue working as they had been.
Failure to support this option or its improper handling will, at worst, cause suboptimal identification of client network location, which is a common occurrence in current Content Delivery Network (CDN) setups.
Contavalli, et al. Informational [Page 7]
RFC 7871 Client Subnet in DNS Queries May 2016
Section 7.1 also provides a mechanism for Stub Resolvers to signal Recursive Resolvers that they do not want ECS treatment for specific queries.
Additionally, operators of Intermediate Nameservers with ECS enabled are allowed to choose how many bits of the address of received queries to forward or to reduce the number of bits forwarded for queries already including an ECS option.
o (Defined in [RFC6891]) OPTION-CODE, 2 octets, for ECS is 8 (0x00 0x08).
o (Defined in [RFC6891]) OPTION-LENGTH, 2 octets, contains the length of the payload (everything after OPTION-LENGTH) in octets.
o FAMILY, 2 octets, indicates the family of the address contained in the option, using address family codes as assigned by IANA in Address Family Numbers [Address_Family_Numbers].
The format of the address part depends on the value of FAMILY. This document only defines the format for FAMILY 1 (IPv4) and FAMILY 2 (IPv6), which are as follows:
o SOURCE PREFIX-LENGTH, an unsigned octet representing the leftmost number of significant bits of ADDRESS to be used for the lookup. In responses, it mirrors the same value as in the queries.
Contavalli, et al. Informational [Page 8]
RFC 7871 Client Subnet in DNS Queries May 2016
o SCOPE PREFIX-LENGTH, an unsigned octet representing the leftmost number of significant bits of ADDRESS that the response covers. In queries, it MUST be set to 0.
o ADDRESS, variable number of octets, contains either an IPv4 or IPv6 address, depending on FAMILY, which MUST be truncated to the number of bits indicated by the SOURCE PREFIX-LENGTH field, padding with 0 bits to pad to the end of the last octet needed.
o A server receiving an ECS option that uses either too few or too many ADDRESS octets, or that has non-zero ADDRESS bits set beyond SOURCE PREFIX-LENGTH, SHOULD return FORMERR to reject the packet, as a signal to the software developer making the request to fix their implementation.
All fields are in network byte order ("big-endian", per [RFC1700], Data Notation).
The ECS option should generally be added by Recursive Resolvers when querying Authoritative Nameservers, as described in Section 12. The option can also be initialized by a Stub Resolver or Forwarding Resolver.
The setup of the ECS option in a Recursive Resolver depends on the client query that triggered the resolution process.
In the usual case, where no ECS option was present in the client query, the Recursive Resolver initializes the option by setting FAMILY of the client's address. It then uses the value of its maximum cacheable prefix length to set SOURCE PREFIX-LENGTH. For privacy reasons, and because the whole IP address is rarely required to determine a tailored response, this length SHOULD be shorter than the full address, as described in Section 11.
If the triggering query included an ECS option itself, it MUST be examined for its SOURCE PREFIX-LENGTH. The Recursive Resolver's outgoing query MUST then set SOURCE PREFIX-LENGTH to the shorter of the incoming query's SOURCE PREFIX-LENGTH or the server's maximum cacheable prefix length.
Contavalli, et al. Informational [Page 9]
RFC 7871 Client Subnet in DNS Queries May 2016
Finally, in both cases, SCOPE PREFIX-LENGTH is set to 0 and ADDRESS is then added up to SOURCE PREFIX-LENGTH number of bits, with trailing 0 bits added, if needed, to fill the final octet. The total number of octets used MUST only be enough to cover SOURCE PREFIX- LENGTH bits, rather than the full width that would normally be used by addresses in FAMILY.
FAMILY and ADDRESS information MAY be used from the ECS option in the incoming query. Passing the existing address data is supportive of the Recursive Resolver being used as the target of a Forwarding Resolver, but could possibly run into policy problems with regard to usage agreements between the Recursive Resolver and Authoritative Nameserver. See Section 12.2 for more discussion on this point. If the Recursive Resolver will not forward FAMILY and ADDRESS data from the incoming ECS option, it SHOULD return a REFUSED response.
Subsequent queries to refresh the data MUST, if unrestricted by an incoming SOURCE PREFIX-LENGTH, specify the longest SOURCE PREFIX- LENGTH that the Recursive Resolver is willing to cache, even if a previous response indicated that a shorter prefix length was sufficient.
A Stub Resolver MAY generate DNS queries with an ECS option that sets SOURCE PREFIX-LENGTH to limit how network information should be revealed. An Intermediate Nameserver that receives such a query MUST NOT make queries that include more bits of client address than in the originating query.
A SOURCE PREFIX-LENGTH value of 0 means that the Recursive Resolver MUST NOT add the client's address information to its queries. The subsequent Recursive Resolver query to the Authoritative Nameserver will then either not include an ECS option or MAY optionally include its own address information, which is what the Authoritative Nameserver will almost certainly use to generate any Tailored Response in lieu of an option. This allows the answer to be handled by the same caching mechanism as other queries, with an explicit indicator of the applicable scope. Subsequent Stub Resolver queries for /0 can then be answered from this cached response.
A Stub Resolver MUST set SCOPE PREFIX-LENGTH to 0. It MAY include FAMILY and ADDRESS data, but should be prepared to handle a REFUSED response if the Intermediate Nameserver that it queries has a policy that denies forwarding of ADDRESS. If there is no ADDRESS set, i.e., SOURCE PREFIX-LENGTH is set to 0, then FAMILY SHOULD be set to the transport over which the query is sent. This is for
Contavalli, et al. Informational [Page 10]
RFC 7871 Client Subnet in DNS Queries May 2016
interoperability; at least one major authoritative server will ignore the option if FAMILY is not 1 or 2, even though it is irrelevant if there are no ADDRESS bits.
Forwarding Resolvers essentially appear to be Stub Resolvers to whatever Recursive Resolver is ultimately handling the query, but they look like a Recursive Resolver to their client. A Forwarding Resolver using this option MUST prepare it as described in Section 7.1.1, "Recursive Resolvers". In particular, a Forwarding Resolver that implements this protocol MUST honor SOURCE PREFIX- LENGTH restrictions indicated in the incoming query from its client. See also Section 7.5.
Since the Recursive Resolver it contacts will treat the Forwarding Resolver like a Stub Resolver, the Recursive Resolver's policies regarding incoming ADDRESS information will apply in the same way. If the Forwarding Resolver receives a REFUSED response when it sends a query that includes a non-zero ADDRESS, it MUST retry with no ADDRESS.
When a query containing an ECS option is received, an Authoritative Nameserver supporting ECS MAY use the address information specified in the option to generate a tailored response.
Authoritative Nameservers that have not implemented or enabled support for the ECS option ought to safely ignore it within incoming queries, per [RFC6891], Section 6.1.2. Such a server MUST NOT include an ECS option within replies to indicate lack of support for it. Implementers of Intermediate Nameservers should be aware, however, that some nameservers incorrectly echo back unknown EDNS0 options. In this protocol, that should be mostly harmless, as the SCOPE PREFIX-LENGTH should come back as 0, thus marking the response as covering all networks.
A query with a wrongly formatted option (e.g., an unknown FAMILY) MUST be rejected and a FORMERR response MUST be returned to the sender, as described in [RFC6891], "Transport Considerations".
An Authoritative Nameserver that implements this protocol and receives an ECS option MUST include an ECS option in its response to indicate that it SHOULD be cached accordingly, regardless of whether the client information was needed to formulate an answer. (Note that
Contavalli, et al. Informational [Page 11]
RFC 7871 Client Subnet in DNS Queries May 2016
the requirement in [RFC6891] to reserve space for the OPT record could mean that the Answer section of the response will be truncated and fall back to TCP indicated accordingly.) If an ECS option was not included in a query, one MUST NOT be included in the response even if the server is providing a Tailored Response -- presumably based on the address from which it received the query.
FAMILY, SOURCE PREFIX-LENGTH, and ADDRESS in the response MUST match those in the query. Echoing back these values helps to mitigate certain attack vectors, as described in Section 11.
SCOPE PREFIX-LENGTH in the response indicates the network for which the answer is intended.
A SCOPE PREFIX-LENGTH value longer than SOURCE PREFIX-LENGTH indicates that the provided prefix length was not specific enough to select the most appropriate Tailored Response. Future queries for the name within the specified network SHOULD use the longer SCOPE PREFIX-LENGTH. Factors affecting whether the Recursive Resolver would use the longer length include the amount of privacy masking the operator wants to provide their users, and the additional resource implications for the cache.
Conversely, a shorter SCOPE PREFIX-LENGTH indicates that more bits than necessary were provided, and the answer is suitable for a broader range of addresses. This could be as short as 0, to indicate that the answer is suitable for all addresses in FAMILY.
As the logical topology of any part of the network with regard to the tailored response can vary, an Authoritative Nameserver may return different values of SCOPE PREFIX-LENGTH for different networks.
Since some queries can result in multiple RRsets being added to the response, there is an unfortunate ambiguity from the original specification as to how SCOPE PREFIX-LENGTH would apply to each individual RRset. For example, multiple types in response to an ANY metaquery could all have different applicable SCOPE PREFIX-LENGTH values, but this protocol only has the ability to signal one. The response SHOULD therefore, include the longest relevant PREFIX-LENGTH of any RRset in the answer, which could have the unfortunate side effect of redundantly caching some data that could be cached more broadly. For the specific case of a Canonical Name (CNAME) chain, the Authoritative Nameserver SHOULD only place the initial CNAME record in the Answer section, to have it cached unambiguously and appropriately. Most modern Recursive Resolvers restart the query with the CNAME, so the remainder of the chain is typically ignored
Contavalli, et al. Informational [Page 12]
RFC 7871 Client Subnet in DNS Queries May 2016
anyway. For message-focused resolvers, rather than RRset-focused ones, this will mean caching the entire CNAME chain at the longest PREFIX-LENGTH of any RRset in the chain.
The specific logic that an Authoritative Nameserver uses to choose a tailored response is not in the scope of this document. Implementers are encouraged, however, to carefully consider their selection of SCOPE PREFIX-LENGTH for the response in the event that the best tailored response cannot be determined, and what the implications would be over the life of the TTL.
Authoritative Nameservers might have situations where one Tailored Response is appropriate for a relatively broad address range, such as an IPv4 /20, except for some exceptions, such as a few /24 ranges within that /20. Because it can't be guaranteed that queries for all longer prefix lengths would arrive before one that would be answered by the shorter prefix length, an Authoritative Nameserver MUST NOT overlap prefixes.
When the Authoritative Nameserver has a longer prefix length Tailored Response within a shorter prefix length Tailored Response, then implementations can either:
1. Deaggregate the shorter prefix response into multiple longer prefix responses, or
2. Alert the operator that the order of queries will determine which answers get cached, and either warn and continue or treat this as an error and refuse to load the configuration.
This choice should be documented for the operator, for example, in the user manual.
When deaggregating to correct the overlap, prefix lengths should be optimized to use the minimum necessary to cover the address space, in order to reduce the overhead that results from having multiple copies of the same answer. As a trivial example, if the Tailored Response for 1.2.0/20 is A but there is one exception of 1.2.3/24 for B, then the Authoritative Nameserver would need to provide Tailored Responses for 1.2.0/23, 1.2.2/24, 1.2.4/22, and 1.2.8/21 all pointing to A, and 1.2.3/24 to B.
When an Intermediate Nameserver uses ECS, whether it passes an ECS option in its own response to its client is predicated on whether the client originally included the option. Because a client that did not use an ECS option might not be able to understand it, the server MUST
Contavalli, et al. Informational [Page 13]
RFC 7871 Client Subnet in DNS Queries May 2016
NOT provide one in its response. If the client query did include the option, the server MUST include one in its response, especially as it could be talking to a Forwarding Resolver, which would need the information for its own caching.
If an Intermediate Nameserver receives a response that has a longer SCOPE PREFIX-LENGTH than SOURCE PREFIX-LENGTH that it provided in its query, it SHOULD still provide the result as the answer to the triggering client request even if the client is in a different address range. The Intermediate Nameserver MAY instead opt to retry with a longer SOURCE PREFIX-LENGTH to get a better reply before responding to its client, as long as it does not exceed a SOURCE PREFIX-LENGTH specified in the query that triggered resolution, but this obviously has implications for the latency of the overall lookup.
The logic for using the cache to determine whether the Intermediate Nameserver already knows the response to provide to its client is covered in the next section.
When an Intermediate Nameserver receives a response containing an ECS option and without the TC bit set, it SHOULD cache the result based on the data in the option. If the TC bit was set, the Intermediate Resolver SHOULD retry the query over TCP to get the complete Answer section for caching.
If FAMILY, SOURCE PREFIX-LENGTH, and SOURCE PREFIX-LENGTH bits of ADDRESS in the response don't match the non-zero fields in the corresponding query, the full response MUST be dropped, as described in Section 11. In a response to a query that specified only SOURCE PREFIX-LENGTH for privacy masking, the FAMILY and ADDRESS fields MUST contain the appropriate non-zero information that the Authoritative Nameserver used to generate the answer, so that it can be cached accordingly.
If no ECS option is contained in the response, the Intermediate Nameserver SHOULD treat this as being equivalent to having received a SCOPE PREFIX-LENGTH of 0, which is an answer suitable for all client addresses. See further discussion on the security implications of this in Section 11.
If a REFUSED response is received from an Authoritative Nameserver, an ECS-aware resolver MUST retry the query without ECS to distinguish the response from one where the Authoritative Nameserver is not responsible for the name, which is a common convention for the REFUSED status. Similarly, a client of a Recursive Resolver SHOULD
Contavalli, et al. Informational [Page 14]
RFC 7871 Client Subnet in DNS Queries May 2016
retry after receiving a REFUSED response because it is not sufficiently clear whether the REFUSED response was because of the ECS option or some other reason.
In the cache, all resource records in the Answer section MUST be tied to the network specified in the response. The appropriate prefix length depends on the relationship between SOURCE PREFIX-LENGTH, SCOPE PREFIX-LENGTH, and the maximum cacheable prefix length configured for the cache.
If SCOPE PREFIX-LENGTH is not longer than SOURCE PREFIX-LENGTH, store SCOPE PREFIX-LENGTH bits of ADDRESS, and then mark the response as valid for all addresses that fall within that range.
Similarly, if SOURCE PREFIX-LENGTH is the maximum configured for the cache, store SOURCE PREFIX-LENGTH bits of ADDRESS, and then mark the response as valid for all addresses that fall within that range.
If SOURCE PREFIX-LENGTH is shorter than the configured maximum and SCOPE PREFIX-LENGTH is longer than SOURCE PREFIX-LENGTH, store SOURCE PREFIX-LENGTH bits of ADDRESS, and then mark the response as valid only to answer client queries that specify exactly the same SOURCE PREFIX-LENGTH in their own ECS option.
The handling of DNSSEC-related records in the Answer section was unspecified in the original draft version of this document and is inconsistently handled in existing implementations. A Resource Record Signature (RRSIG) must obviously be tied to the RRset that it signs, but it is RECOMMENDED that all other DNSSEC records be scoped at /0. See Section 9 for more information.
Note that the Additional and Authority sections from a DNS response message are specifically excluded here. Any records from these sections MUST NOT be tied to a network. See Section 7.4 for more information.
Records that are cached as /0 because of a query's SOURCE PREFIX- LENGTH of 0 MUST be distinguished from those that are cached as /0 because of a response's SCOPE PREFIX-LENGTH of 0. The former should only be used for other /0 queries that the Intermediate Resolver receives, but the latter is suitable as a response for all networks.
Contavalli, et al. Informational [Page 15]
RFC 7871 Client Subnet in DNS Queries May 2016
Although omitting network-specific caching will significantly simplify an implementation, the resulting drop in cache hits is very likely to defeat most latency benefits provided by ECS. Therefore, implementing full caching support as described in this section is strongly RECOMMENDED.
Enabling support for ECS in an Intermediate Nameserver will significantly increase the size of the cache, reduce the number of results that can be served from cache, and increase the load on the server. Implementing the mitigation techniques described in Section 11 is strongly recommended. For cache size issues, implementers should consider data storage formats that allow the same answer data to be shared among multiple prefixes.
Cache lookups are first done as usual for a DNS query, using the query tuple of <name, type, class>. Then, the appropriate RRset MUST be chosen based on the longest prefix matching. The client address to use for comparison will depend on whether the Intermediate Nameserver received an ECS option in its client query.
o If no ECS option was provided, the client's address is used.
o If there was an ECS option specifying SOURCE PREFIX-LENGTH and ADDRESS covering the client's address, the client address is used but SOURCE PREFIX-LENGTH is initially ignored. If no covering entry is found and SOURCE PREFIX-LENGTH is shorter than the configured maximum length allowed for the cache, repeat the cache lookup for an entry that exactly matches SOURCE PREFIX-LENGTH. These special entries, which do not cover longer prefix lengths, occur as described in the previous section.
o If there was an ECS option with an ADDRESS, the ADDRESS from it MAY be used if the local policy allows. The policy can vary depending on the agreements the operator of the Intermediate Nameserver has with Authoritative Nameserver operators; see Section 12.2. If the policy does not allow it, a REFUSED response SHOULD be sent. See Section 7.5 for more information.
If a matching network is found and the relevant data is unexpired, the response is generated as per Section 7.2.
If no matching network is found, the Intermediate Nameserver MUST perform resolution as usual. This is necessary to avoid Tailored Responses in the cache from being returned to the wrong clients, and
Contavalli, et al. Informational [Page 16]
RFC 7871 Client Subnet in DNS Queries May 2016
to avoid a single query coming from a client on a different network from polluting the cache with a Tailored Response for all the users of that resolver.
The prohibition against tying ECS data to records from the Authority and Additional sections left an unfortunate ambiguity in the original specification, primarily with regard to negative answers. The expectation of the original authors was that ECS would only really be used for address requests and the positive result in the response's Answer section, which was the use case that was driving the definition of the protocol.
For negative answers, some independent implementations of both resolvers and authorities did not see the section restriction as necessarily meaning that a given name and type must only have either positive ECS-tagged answers or a negative answer. They support being able to tell one part of the network that the data does not exist, while telling another part of the network that it does.
Several other implementations, however, do not support being able to mix positive and negative answers; thus, interoperability is a problem. It is RECOMMENDED that no specific behavior regarding negative answers be relied upon, but that Authoritative Nameservers should conservatively expect that Intermediate Nameservers will treat all negative answers as /0; therefore, they SHOULD set SCOPE PREFIX- LENGTH accordingly.
This issue is expected to be revisited in a future revision of the protocol, possibly blessing the mixing of positive and negative answers. There are implications for cache data structures that developers should consider when writing new ECS code.
The delegations case is a bit easier to tease out. In operational practice, if an authoritative server is using address information to provide customized delegations, it is the resolver that will be using the answer for its next iterative query. Addresses in the Additional section SHOULD therefore ignore ECS data, and the Authoritative Nameserver SHOULD return a zero SCOPE PREFIX-LENGTH on delegations. A Recursive Resolver SHOULD treat a non-zero SCOPE PREFIX LENGTH in a delegation as though it were zero.
Generally, ECS options will only be present in DNS messages between a Recursive Resolver and an Authoritative Nameserver, i.e., one hop. However, in certain configurations, for example, multi-tier nameserver setups, it may be necessary to implement transitive behavior on Intermediate Nameservers.
Any Intermediate Nameserver that forwards ECS options received from its clients MUST fully implement the caching behavior described in Section 7.3.
An Intermediate Nameserver MAY forward ECS options with address information. This information MAY match the source IP address of the incoming query, and MAY have more or fewer address bits than the nameserver would normally include in a locally originated ECS option. If an Intermediate Nameserver receives a query with SOURCE PREFIX- LENGTH set to 0, it MUST NOT include client address information in queries made to resolve that client's request (see Section 7.1.2).
If, for any reason, the Intermediate Nameserver does not want to use the information in an ECS option it receives (too little address information, network address from a range not authorized to use the server, private/unroutable address space, etc.), it SHOULD drop the query and return a REFUSED response. Note again that a query MUST NOT be refused solely because it provides 0 address bits.
Be aware that at least one major existing implementation does not return REFUSED and instead just processes the query as though the problematic information were not present. This can lead to anomalous situations, such as a response from the Intermediate Nameserver that indicates it is tailored for one network (the one passed in the original query, since the ADDRESS must match) when actually it is for another network (the one which contains the address that the Intermediate Nameserver saw as making the query).
The presence or absence of an EDNS0 OPT resource record ([RFC6891]) containing an ECS option in a DNS query does not change the usage of the resource records and mechanisms used to provide data origin authentication and data integrity to the DNS, as described in [RFC4033], [RFC4034], and [RFC4035]. OPT records are not signed.
Use of this option, however, does imply increased DNS traffic between any given Recursive Resolver and Authoritative Nameserver, which could be another barrier to further DNSSEC adoption in this area.
The initial version of this protocol, against which several Authoritative and Recursive Nameserver implementations were written, did not discuss the handling of DNSSEC RRs; thus, it is expected that there are operational inconsistencies in handling them.
Given the intention of this document to describe how ECS is currently deployed, specifying new requirements for DNSSEC handling is out of scope. However, some recommendations can be made as to what is most likely to result in successful interoperation for a DNSSEC-signed ECS zone, mainly from the point of view of Authoritative Nameservers.
Most DNSSEC records SHOULD be scoped at /0, except for the RRSIG records, which MUST be tied to the RRset that they sign in a Tailored Response. While it is possible to conceive of a way to get other DNSSEC records working in a network-specific way, it has little apparent benefit or likelihood of working with deployed validating resolvers.
One further implication here is that, despite the discussion about negative answers in Section 7.4, scoping NextSECure (NSEC) or NSEC3 records at /0 per the previous paragraph necessarily implies that DNSSEC-signed negative answers must also be network-invariant.
Special awareness of ECS in devices that perform Network Address Translation (NAT) as described in [RFC2663] is not required; queries can be passed through as is. The client's network address SHOULD NOT be added, and existing ECS options, if present, SHOULD NOT be modified by NAT devices.
In large-scale global networks behind a NAT device (but, for example with Centralized Resolver infrastructure), an internal Intermediate Nameserver might have detailed network layout information, and may
Contavalli, et al. Informational [Page 19]
RFC 7871 Client Subnet in DNS Queries May 2016
know which external subnets are used for egress traffic by each internal network. In such cases, the Intermediate Nameserver MAY use that information when originating ECS options.
In other cases, if a Recursive Resolver knows that it is situated behind a NAT device, it SHOULD NOT originate ECS options with their external IP address and instead rely on downstream Intermediate Nameservers to do so. It MAY, however, choose to include the option with their internal address for the purposes of signaling its own limit for SOURCE PREFIX-LENGTH.
Full treatment of special network addresses is beyond the scope of this document; handling them will likely differ according to the operational environments of each service provider. As a general guideline, if an Authoritative Nameserver on the publicly routed Internet receives a query that specifies an ADDRESS in [RFC1918] or [RFC4193] private address space, it SHOULD ignore ADDRESS and look up its answer based on the address of the Recursive Resolver. In the response, it SHOULD set SCOPE PREFIX-LENGTH to cover all of the relevant private space. For example, a query for ADDRESS 10.1.2.0 with a SOURCE PREFIX-LENGTH of 24 would get a returned SCOPE PREFIX- LENGTH of 8. The Intermediate Nameserver MAY elect to cache the answer under one entry for special-purpose addresses [RFC6890]; see Section 11.3 of this document.
With the ECS option, the network address of the client that initiated the resolution becomes visible to all servers involved in the resolution process. Additionally, it will be visible from any network traversed by the DNS packets.
To protect users' privacy, Recursive Resolvers are strongly encouraged to conceal part of the user's IP address by truncating IPv4 addresses to 24 bits. 56 bits are recommended for IPv6, based on [RFC6177].
ISPs should have more detailed knowledge of their own networks. That is, they might know that all 24-bit prefixes in a /20 are in the same area. In those cases, for optimal cache utilization and improved privacy, the ISP's Recursive Resolver SHOULD truncate IP addresses in this /20 to just 20 bits, instead of 24 as recommended above.
Users who wish their full IP address to be hidden need to configure their client software, if possible, to include an ECS option specifying the wildcard address (i.e., a SOURCE PREFIX-LENGTH of 0).
Contavalli, et al. Informational [Page 20]
RFC 7871 Client Subnet in DNS Queries May 2016
As described in previous sections, this option will be forwarded across all the Recursive Resolvers supporting ECS, which MUST NOT modify it to include the network address of the client.
Note that even without an ECS option, any server queried directly by the user will be able to see the full client IP address. Recursive Resolvers or Authoritative Nameservers MAY use the source IP address of queries to return a cached entry or to generate a Tailored Response that best matches the query.
ECS adds information to the DNS query tuple (q-tuple). This allows an attacker to send a caching Intermediate Nameserver multiple queries with spoofed IP addresses either in the ECS option or as the source IP. These queries will trigger multiple outgoing queries with the same name, type, and class, just with different address information in the ECS option.
With multiple queries for the same name in flight, the attacker has a higher chance of success to send a matching response with SCOPE PREFIX-LENGTH set to 0 to get it cached for all hosts.
To counter this, the ECS option in a response packet MUST contain the full FAMILY, ADDRESS, and SOURCE PREFIX-LENGTH fields from the corresponding query. Intermediate Nameservers processing a response MUST verify that these match, and they SHOULD discard the entire response if they do not.
The requirement to discard is categorized as "SHOULD" instead of "MUST" because it stands in opposition to the instruction in Section 7.3, which states that a response lacking an ECS option should be treated as though it had one of SCOPE PREFIX-LENGTH of 0. If that is always true, then an attacker does not need to worry about matching the original ECS option data and just needs to flood back responses that have no ECS option at all.
This type of attack could be detected in ongoing operations by marking whether the responding nameserver had previously been sending ECS options and/or by taking note of an incoming flood of bogus responses and flagging the relevant query for re-resolution. This type of detection is more complex than existing nameserver responses to spoof floods, and it would also need to be sensitive to a nameserver legitimately stopping ECS replies even though it had previously given them.
It is simple for an arbitrary resolver or client to provide false information in the ECS option, or to send UDP packets with forged source IP addresses.
This could be used to:
o pollute the cache of Intermediate Resolvers by filling it with results that will rarely (if ever) be used.
o reverse-engineer the algorithms (or data) used by the Authoritative Nameserver to calculate Tailored Responses.
o mount a denial-of-service attack against an Intermediate Nameserver by forcing it to perform many more recursive queries than it would normally do, due to how caching is handled for queries containing the ECS option.
Even without malicious intent, Centralized Resolvers providing answers to clients in multiple networks will need to cache different responses for different networks, putting more memory pressure on the cache.
To mitigate those problems:
o Recursive Resolvers implementing ECS should only enable it in deployments where it is expected to bring clear advantages to the end users, such as when expecting clients from a variety of networks or from a wide geographical area. Due to the high cache pressure introduced by ECS, the feature SHOULD be disabled in all default configurations.
o Recursive Resolvers SHOULD limit the number of networks and answers they keep in the cache for any given query.
o Recursive Resolvers SHOULD limit the total number of different networks that they keep in cache.
o Recursive Resolvers MUST NOT send an ECS option with SOURCE PREFIX-LENGTH providing more bits in ADDRESS than they are willing to cache responses for.
o Recursive Resolvers should implement algorithms to improve the cache hit rate, given the size constraints indicated above. Recursive Resolvers MAY, for example, decide to discard more- specific cache entries first.
Contavalli, et al. Informational [Page 22]
RFC 7871 Client Subnet in DNS Queries May 2016
o Authoritative Nameservers and Recursive Resolvers should discard ECS options that are either obviously forged or otherwise known to be wrong. They SHOULD at least treat unroutable addresses, such as some of the address blocks defined in [RFC6890], as equivalent to the Recursive Resolver's own identity. They SHOULD ignore and never forward ECS options specifying other routable addresses that are known not to be served by the query source.
o The ECS option is just a hint to Authoritative Nameservers for customizing results. They can decide to ignore the content of the ECS option based on blacklists or whitelists, rate-limiting mechanisms, or any other logic implemented in the software.
When implementing a Recursive Resolver, there are two strategies on deciding when to include an ECS option in a query. At this stage, it's not clear which strategy is best.
A Recursive Resolver can send the ECS option with every outgoing query. However, it is RECOMMENDED that resolvers remember which Authoritative Nameservers did not return the option with their response and omit client address information from subsequent queries to those nameservers.
Additionally, Recursive Resolvers SHOULD be configured never to send the option when querying root, top-level, and effective top-level (i.e., "public suffix" [Public_Suffix_List]) domain servers. These domains are delegation-centric and are very unlikely to generate different responses based on the address of the client.
When probing, it is important that several things are probed: support for ECS, support for EDNS0, support for EDNS0 options, or possibly an unreachable nameserver. Various implementations are known to drop DNS packets with OPT RRs (with or without options), thus several probes are required to discover what is supported.
Probing, if implemented, MUST be repeated periodically, e.g., daily. If an Authoritative Nameserver indicates ECS support for one zone, it is to be expected that the nameserver supports ECS for all of its zones. Likewise, an Authoritative Nameserver that uses ECS information for one of its zones MUST indicate support for the option in all of its responses to ECS queries. If the option is supported but not actually used for generating a response, its SCOPE PREFIX- LENGTH MUST be set to 0.
As described previously, it is expected that only a few Recursive Resolvers will need to use ECS, and that it will generally be enabled only if it offers a clear benefit to the users.
To avoid the complexity of implementing a probing and detection mechanism (and the possible query loss/delay that may come with it), an implementation could use a whitelist of Authoritative Nameservers to send the option to, likely specified by their domain name. Implementations MAY also allow additional configuring of this based on other criteria, such as zone or query type. As of the time of this writing, at least one implementation makes use of a whitelist.
An advantage of using a whitelist is that partial client address information is only disclosed to nameservers that are known to use the information, improving privacy.
A drawback is scalability. The operator needs to track which Authoritative Nameservers support ECS, making it harder for new Authoritative Nameservers to start using the option.
Similarly, Authoritative Nameservers can also use whitelists to limit the feature to only certain clients. For example, a CDN that does not want all of their mapping trivially walked might require a legal agreement with the Recursive Resolver operator, to clearly describe the acceptable use of the feature.
The maintenance of access control mechanisms is out of scope for this protocol definition.
1. A Stub Resolver, SR, with the IP address 2001:0db8:fd13:4231:2112:8a2e:c37b:7334 tries to resolve www.example.com by forwarding the query to the Recursive Resolver, RNS, asking for recursion.
2. RNS, supporting ECS, looks up www.example.com in its cache. An entry is found neither for www.example.com nor for example.com.
3. RNS builds a query to send to the root and .com servers. The implementation of RNS provides facilities so that an administrator can configure it not to forward ECS in certain cases. In particular, RNS is configured not to include an ECS option when talking to Top-Level-Domain or root nameservers, as described in Section 7.1. Thus, no ECS option is added, and resolution is performed as usual.
Contavalli, et al. Informational [Page 24]
RFC 7871 Client Subnet in DNS Queries May 2016
4. RNS now knows the next server to query: the Authoritative Nameserver, ANS, responsible for example.com.
5. RNS prepares a new query for www.example.com, including an ECS option with:
* OPTION-CODE set to 8.
* OPTION-LENGTH set to 0x00 0x0b for the following fixed 4 octets plus the 7 octets that will be used for ADDRESS.
* FAMILY set to 0x00 0x02, as IP is an IPv6 address.
* SOURCE PREFIX-LENGTH set to 0x38, as RNS is configured to conceal the last 72 bits of every IPv6 address.
* SCOPE PREFIX-LENGTH set to 0x00, as specified by this document for all queries.
* ADDRESS set to 0x20 0x01 0x0d 0xb8 0xfd 0x13 0x42, providing only the first 56 bits of the IPv6 address.
6. The query is sent. ANS understands and uses ECS. It parses the ECS option, and generates a Tailored Response.
7. Due its internal implementation, ANS finds a response that is tailored for the whole /16 of the client that performed the query.
8. ANS adds an ECS option in the response, containing:
* OPTION-CODE set to 8.
* OPTION-LENGTH set to 0x00 0x07.
* FAMILY set to 0x00 0x02.
* SOURCE PREFIX-LENGTH set to 0x38, copied from the query.
* SCOPE PREFIX-LENGTH set to 0x30, indicating a /48 network.
* ADDRESS set to 0x20 0x01 0x0d 0xb8 0xfd 0x13 0x42, copied from the query.
9. RNS receives the response containing an ECS option. It verifies that FAMILY, SOURCE PREFIX-LENGTH, and ADDRESS match the query. If not, the message is discarded.
Contavalli, et al. Informational [Page 25]
RFC 7871 Client Subnet in DNS Queries May 2016
10. The response is interpreted as usual. Since the response contains an ECS option, ADDRESS, SCOPE PREFIX-LENGTH, and FAMILY in the response are used to cache the entry.
11. RNS sends a response to Stub Resolver, SR, without including an ECS option.
12. RNS receives another query to resolve www.example.com. This time, a response is cached. The response, however, is tied to a particular network. If the client's address matches any network in the cache, then the response is returned from the cache. Otherwise, another query is performed. If multiple results match, the one with the longest SCOPE PREFIX-LENGTH is chosen, as per common best-network-match algorithms.
[VANDERGAAST] Contavalli, C., Gaast, W., Leach, S., and E. Lewis, "Client Subnet in DNS Requests", Work in Progress, draft-vandergaast-edns-client-subnet-02, July 2013.
Acknowledgements
The authors wish to thank Darryl Rodden for his work as a co-author, and the following people for reviewing this document and for providing useful feedback: Paul S. R. Chisholm, B. Narendran, Leonidas Kontothanassis, David Presotto, Philip Rowlands, Chris Morrow, Kara Moscoe, Alex Nizhner, Warren Kumari, and Richard Rabbat from Google; Terry Farmer, Mark Teodoro, Edward Lewis, and Eric Burger from Neustar; David Ulevitch and Matthew Dempsky from OpenDNS; Patrick W. Gilmore and Steve Hill from Akamai; Colm MacCarthaigh and Richard Sheehan from Amazon; Tatuya Jinmei from Infoblox; Andrew Sullivan from Dyn; John Dickinson from Sinodun; Mark Delany from Apple; Yuri Schaeffer from NLnet Labs; Duane Wessels Verisign; Antonio Querubin; Daniel Kahn Gillmor from the ACLU; Evan Hunt and Mukund Sivaraman from the Internet Software Consortium; Russ Housley from Vigilsec; Stephen Farrell from Trinity College Dublin; Alissa Cooper from Cisco; Suzanne Woolf; and all of the other people that replied to our emails on various mailing lists.
Contavalli, et al. Informational [Page 28]
RFC 7871 Client Subnet in DNS Queries May 2016
Contributors
The individuals below contributed significantly to this document.
Edward Lewis ICANN 12025 Waterfront Drive, Suite 300 Los Angeles, CA 90094-2536 United States
Email: edward.lewis@icann.org
Sean Leach Fastly P.O. Box 78266 San Francisco, CA 94107 United States
Jason Moreau Akamai Technologies 150 Broadway Cambridge, MA 02142-1413 United States
Contavalli, et al. Informational [Page 29]
RFC 7871 Client Subnet in DNS Queries May 2016
Authors' Addresses
Carlo Contavalli Google 1600 Amphitheater Parkway Mountain View, CA 94043 United States
Email: ccontavalli@google.com
Wilmer van der Gaast Google Belgrave House, 76 Buckingham Palace Road London SW1W 9TQ United Kingdom
Email: wilmer@google.com
David C Lawrence Akamai Technologies 150 Broadway Cambridge, MA 02142-1054 United States
Email: tale@akamai.com
Warren Kumari Google 1600 Amphitheatre Parkway Mountain View, CA 94043 United States