| RFC 8801 | Provisioning Domains | July 2020 |
| Pfister, et al. | Standards Track | [Page] |
Provisioning Domains (PvDs) are defined as consistentsets of network configuration information. PvDs allows hosts to manageconnections to multiple networks and interfaces simultaneously, such aswhen a home router provides connectivity through both a broadband andcellular network provider.¶
This document defines a mechanism for explicitly identifying PvDs througha Router Advertisement (RA) option. This RA option announces a PvD identifier,which hosts can compare to differentiate between PvDs. The option can directlycarry some information about a PvD and can optionally point toPvD Additional Information that can be retrieved using HTTP over TLS.¶
This is an Internet Standards Track document.¶
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). Further information on Internet Standards is available in Section 2 of RFC 7841.¶
Information about the current status of this document, any errata, and how to provide feedback on it may be obtained athttps://www.rfc-editor.org/info/rfc8801.¶
Copyright (c) 2020 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 (https://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.¶
Provisioning Domains (PvDs) are defined in[RFC7556] as consistentsets of network configuration information. This information includesproperties that are traditionally associated with a single networkinginterface, such as source addresses, DNS configuration, proxy configuration,and gateway addresses.¶
Clients that are aware of PvDs can take advantage of multiple networkinterfaces simultaneously. This enables using two PvDs in parallel forseparate connections or for multi-path transports.¶
While most PvDs today are discovered implicitly (such as by receivinginformation via Router Advertisements from a router on a networkthat a client host directly connects to),[RFC7556] also defines the notionof Explicit PvDs. IPsec Virtual Private Networks are considered Explicit PvDs,but Explicit PvDs can also be discovered via the local network router.Discovering Explicit PvDs allows two key advancements in managing multiple PvDs:¶
While[RFC7556] defines the concept of Explicit PvDs, it does not definethe mechanism for discovering multiple Explicit PvDs on a single networkand their Additional Information.¶
This document specifies a way to identify PvDs with Fully QualifiedDomain Names (FQDNs), called PvD IDs. Those identifiers are advertised ina new Router Advertisement (RA)[RFC4861]option calledthe PvD Option, which, when present, associatesthe PvD ID with all the information present in the Router Advertisementas well as any configuration object, such as addresses, derived fromit. The PvD Option may also contain a set ofother RA options, along with an optional inner Router Advertisementmessage header. These options and optional inner header are only visibleto 'PvD-aware' hosts, allowing such hosts to have a specialized view of thenetwork configuration.¶
Since PvD IDs are used to identify different ways to access theInternet, multiple PvDs (with different PvD IDs) can be provisioned ona single host interface. Similarly, the same PvD ID could be used ondifferent interfaces of a host in order to inform that those PvDsultimately provide equivalent services.¶
This document also introduces a mechanism for hosts to retrieve optional Additional Information related to a specific PvD by means of an HTTP-over-TLS query using a URI derived from the PvD ID. The retrieved JSON object contains Additional Information that would typically be considered too large to be directly included in the Router Advertisement but might be considered useful to the applications, or even sometimes users, when choosing which PvD should be used.¶
For example, if Alice has both a cellular network provider and abroadband provider in her home, her PvD-aware devices and applicationswould be aware of both available uplinks. These applicationscould fail-over between these networks or run connections over both(potentially using multi-path transports). Applications could also selectspecific uplinks based on the properties of the network; for example,if the cellular network provides free high-quality video streaming,a video-streaming application could select that network while most of theother traffic on Alice's device uses the broadband provider.¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14[RFC2119][RFC8174] when, and only when, they appear in all capitals, as shown here.¶
This document uses the following terminology:¶
Explicit PvDs are identified by a PvD ID. The PvD ID is a FullyQualified Domain Name (FQDN) that identifies the network operator.Network operatorsMUST use names that they own or manage toavoid naming conflicts. The same PvD IDMAY be used inseveral access networks when they ultimately provide identical services(e.g., in all home networks subscribed to the same service); else, thePvD IDMUST be different to followSection 2.4 of [RFC7556].¶
This document introduces a Router Advertisement (RA) option calledthe PvD Option. It is used to convey the FQDN identifying a given PvD (seeFigure 1), bind the PvD ID with configurationinformation received over DHCPv4 (seeSection 3.4.2), enablethe use of HTTP over TLS to retrieve the PvD Additional InformationJSON object (seeSection 4), as well as containany otherRA options that would otherwise be valid in the RA.¶
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+| Type | Length |H|L|R| Reserved | Delay |+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+| Sequence Number | ...+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ ...... PvD ID FQDN ...... +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+... | Padding |+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+| ...... Router Advertisement message header ...... (Only present when R-flag is set) ...... |+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+| Options ...+-+-+-+-+-+-+-+-+-+-+-+-
Figure 2 shows an example of a PvD Option with "example.org" as the PvD ID FQDN and includes both a Recursive DNS Server (RDNSS) option and a Prefix Information Option. It has a Sequence Number of 123 and indicates the presence of PvD Additional Information that is expected to be fetched with a delay factor of 1.¶
0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1+---------------+-----------------------------------------------+| Type: 21 | Length: 12 |1|0|0| Reserved |Delay:1|+---------------+-------------------------------+---------------+| Seq number: 123 | 7 | e |+---------------+-----------------------------------------------+| x | a | m | p |+---------------------------------------------------------------+| l | e | 3 | o |+---------------------------------------------------------------+| r | g | 0 | 0 (padding) |+---------------------------------------------------------------+| 0 (padding) | 0 (padding) | 0 (padding) | 0 (padding) |+---------------+---------------+---------------+---------------+| RDNSS option (RFC 8106) length: 5 ...... ...... |+---------------------------------------------------------------+| Prefix Information Option (RFC 4861) length: 4 ...... |... |+---------------------------------------------------------------+
A routerMAY send RAs containing one PvD Option butMUST NOT include more than one PvD Option in each RA. The PvD OptionMUST NOT contain further PvD Options.¶
The PvD OptionMAY contain zero, one, or more RA options that would otherwise be valid as part of the same RA. Such options are processed by PvD-aware hosts and ignored by other hosts as perSection 4.2 of [RFC4861].¶
In order to provide multiple different PvDs, a routerMUST send multiple RAs. RAs sent from different link-local source addresses establish distinct Implicit PvDs in the absence of a PvD Option. Explicit PvDsMAY share link-local source addresses with an Implicit PvD and any number of other Explicit PvDs.¶
In other words, different Explicit PvDsMAY be advertised with RAs using the same link-local source address, but different Implicit PvDs, advertised by different RAs,MUST use different link-local addresses because these Implicit PvDs are identified by the source addresses of the RAs. If a link-local address on the router is changed, then any new RA will be interpreted as a different Implicit PvD by PvD-aware hosts.¶
As specified in[RFC4861] and[RFC6980], when the set of options causes the size of an advertisement to exceed the link MTU, multiple router advertisementsMUST be sent to avoid fragmentation, each containing a subset of the options. In such cases, the PvD Option header (i.e., all fields except the Options field)MUST be repeated in all the transmitted RAs. The options within the Options fieldMAY be transmitted onlyonce, included in one of the transmitted PvD Options.¶
As the PvD Option has a new option code, non-PvD-aware hosts will simply ignore the PvD Option and all the options it contains (seeSection 4.2 of [RFC4861]). This ensures the backward compatibility required inSection 3.3 of [RFC7556]. This behavior allows for a mixed-mode network where a mix of PvD-aware and non-PvD-aware hosts coexist.¶
HostsMUST associate received RAs and included configuration information (e.g., Router Valid Lifetime, Prefix Information[RFC4861], Recursive DNS Server[RFC8106], and Routing Information[RFC4191] options) with the Explicit PvD identified by the first PvD Option present in the received RA, if any, or with the Implicit PvD identified by the host interface and the source address of the received RA otherwise. If an RA message header is present both within the PvD Option and outside it, the header within the PvD Option takes precedence.¶
In case multiple PvD Options are found in a given RA, hostsMUST ignore all but the first PvD Option.¶
If a host receives PvD Options flags that it does not recognize (currently in the Reserved field), itMUST ignore these flags.¶
Similarly, hostsMUST associate all network configuration objects (e.g., default routers, addresses, more specific routes, and DNS Recursive Resolvers) with the PvD associated with the RA that provisioned the object. For example, addresses that are generated using a received Prefix Information Option (PIO) are associated with the PvD of the last received RA that included the given PIO.¶
PvD IDsMUST be compared in a case-insensitive manner as defined by[RFC4343]. For example, "pvd.example.com." or"PvD.Example.coM."would refer to the same PvD.¶
While performing PvD-specific operations such as resolving names, executing the default address selection algorithm[RFC6724], or executing the default router selection algorithm when forwarding packets[RFC4861][RFC4191][RFC8028], hosts and applicationsMAY consider only the configuration associated with any non-empty subset of PvDs. For example, a hostMAY associate a given process with a specific PvD, or a specific set of PvDs, while associating another process with another PvD. A PvD-aware application might also be able to select, on a per-connection basis, which PvDs should be used. In particular, constrained devices such as small battery-operated devices (e.g., Internet of Things (IoT)) or devices with limited CPU or memory resources may purposefully use a single PvD while ignoring some received RAs containing different PvD IDs.¶
The way an application expresses its desire to use a given PvD, or a set of PvDs, and the way this selection is enforced are out of the scope of this document. Useful insights about these considerations can be found in[MPVD-API].¶
When a host retrieves stateless configuration elements using DHCPv6 (e.g., DNS recursive resolvers or DNS domain search lists[RFC3646]), theyMUST be associated with all the Explicit and Implicit PvDs received on the same interface and contained in an RA with the O-flag set[RFC4861].¶
When a host retrieves stateful assignments using DHCPv6, suchassignmentsMUST be associated with the received PvD that wasreceived with RAs with the M-flag set and including a matching PIO.A PIO is considered to match a DHCPv6 assignment when the IPv6 prefixfrom the PIO includes the assignment from DHCPv6. For example,if a PvD's associated PIO defines the prefix2001:db8:cafe::/64,a DHCPv6 IA_NA message that assigns the address2001:db8:cafe::1234:4567would be considered to match.¶
In cases where an address would be assigned by DHCPv6 and no matchingPvD could be found, hostsMAY associate the assigned addresswith anyImplicit PvD received on the same interface or to multiple Implicit PvDsreceived on the same interface. This is intended to resolvebackward-compatibilityissues with rare deployments choosing to assign addresses with DHCPv6 whilenot sending any matching PIO. Implementations are suggested to flag or logsuch scenarios as errors to help detect misconfigurations.¶
Associating DHCPv4[RFC2131] configuration elements with Explicit PvDs allows hosts to treat a set of IPv4 and IPv6 configurations as a single PvD with shared properties. For example, consider a router that provides two different uplinks. One could be a broadband network that has data rate and streaming properties described in PvD Additional Information and that provides both IPv4 and IPv6 network access. The other could be a cellular network that provides only IPv6 network access and uses NAT64[RFC6146]. The broadband network can be represented by an Explicit PvD that points to the Additional Information and also marks association with DHCPv4 information. The cellular network can be represented by a different Explicit PvD that is not associated with DHCPv4.¶
When a PvD-aware host retrieves configuration elements from DHCPv4, the information is associated either with a single Explicit PvD on that interface or else with all Implicit PvDs on the same interface.¶
An Explicit PvD indicates its association with DHCPv4 information by setting the L-flag in the PvD Option. If there is exactly one Explicit PvD that sets this flag, hostsMUST associate the DHCPv4 information with that PvD. Multiple Explicit PvDs on the same interface marking this flag is a misconfiguration, and hostsSHOULD NOT associate the DHCPv4 information with any Explicit PvD in this case.¶
If no single Explicit PvD claims association with DHCPv4, the configuration elements coming from DHCPv4MUST be associated with all Implicit PvDs identified by the interface on which the DHCPv4 transaction happened. This maintains existing host behavior.¶
The situation in which a host shares connectivity from an upstream interface (e.g., cellular) to a downstream interface (e.g., Wi-Fi) is known as 'tethering'. Techniques such as ND Proxy[RFC4389], 64share[RFC7278], or prefix delegation (e.g., using DHCPv6-PD[RFC8415]) may be used for that purpose.¶
Whenever the RAs received from the upstream interface contain a PvD Option, hosts that are sharing connectivitySHOULD include a PvD Option within the RAs sent downstream with:¶
The values of the R-flag, Router Advertisement messageheader, and Options field depend on whether or not the connectivity shouldbe shared only with PvD-aware hosts (seeSection 3.2). In particular,all options received within the upstream PvD Option and included inthe downstream RASHOULD be included in the downstream PvD Option.¶
PvD-aware hosts can be provisioned with recursive DNS servers viaRA options passed within an Explicit PvD, via RA options associatedwith an Implicit PvD, via DHCPv6 or DHCPv4, or from some otherprovisioning mechanism that creates an Explicit PvD (such as a VPN).In all of these cases, the recursive DNS server addressesSHOULD beassociated with the corresponding PvD. Specifically, queries sentto a configured recursive DNS serverSHOULD be sent from alocal IPaddress that was provisioned for the PvD via RA or DHCP. Answersreceived from the DNS serverSHOULD only be used on the same PvD.¶
PvD-aware applications will be able to select which PvD(s) to use for DNS resolution and connections, which allows them to effectively use multiple Explicit PvDs. In order to support non-PvD-aware applications, however, PvD-aware hostsSHOULD ensure that non-PvD-aware name resolution APIs like "getaddrinfo" only use resolvers from a single PvD for a given query. Handling DNS across PvDs is discussed inSection 5.2.1 of [RFC7556], and PvD APIs are discussed inSection 6 of [RFC7556].¶
Maintaining the correct usage of DNS within PvDs avoids variouspractical errors such as:¶
Additional information about the network characteristics can be retrieved based on the PvD ID. This set of information is called PvD Additional Information and is encoded as a JSON object[RFC8259]. This JSON object is restricted to the Internet JSON (I-JSON) profile, as defined in[RFC7493].¶
The purpose of this JSON object is to provide Additional Information to applications on a client host about the connectivity that is provided using a given interface and source address. It typically includes data that would be considered too large, or not critical enough, to be provided within an RA option. The information contained in this objectMAY be used by the operating system, network libraries, applications, or users in order to decide which set of PvDs should be used for which connection, as described inSection 3.4.¶
The Additional Information related to a PvD is specifically intended to be optional and is targeted at optimizing or informing the behavior of user-facing hosts. This information can be extended to provide hints for host system behavior (such as captive portal or walled-garden PvD detection) or application behavior (describing application-specific services offered on a given PvD). This content may not be appropriate for light-weight IoT devices. IoT devices might need only a subset of the information and would in some cases prefer a smaller representation like Concise Binary Object Representation (CBOR)[RFC7049]. Delivering a reduced version of the PvD Additional Information designed for such devices is not defined in this document.¶
When the H-flag of the PvD Option is set, hostsMAY attempt to retrieve the PvD Additional Information associated with a given PvD by performing an HTTP-over-TLS[RFC2818] GET query tohttps://<PvD-ID>/.well-known/pvd. Inversely, hostsMUST NOT do so whenever the H-flag is not set.¶
Recommendations for how to use TLS securely can be found in[RFC7525].¶
When a host retrieves the PvD Additional Information, itMUSTverify that the TLS server certificate is valid for the performedrequest, specifically, that a DNS-ID[RFC6125]on the certificate is equal tothe PvD ID expressed as an FQDN. This validation indicates that theowner of the FQDN authorizes its use with the prefix advertised by the router.If this validation fails, hostsMUST close the connection andtreat the PvDas if it has no Additional Information.¶
HTTP requests and responses for PvD Additional Information use the"application/pvd+json" media type (seeSection 8.5). ClientsSHOULD include this media type as an Accept header field intheir GETrequests, and serversMUST mark this media type as theirContent-Typeheader field in responses.¶
Note that the DNS name resolution of the PvD ID, any connections madefor certificate validation (such as Online Certificate Status Protocol (OCSP)[RFC6960]), andthe HTTP request itselfMUST be performed using the consideredPvD.In other words, the name resolution, PKI checks, source addressselection, as well as the next-hop router selectionMUST beperformedwhile exclusively using the set of configuration information attachedwith the PvD, as defined inSection 3.4. In somecases, itmay therefore be necessary to wait for an address to be available foruse (e.g., once the Duplicate Address Detection or DHCPv6 processesare complete) before initiating the HTTP-over-TLS query. In order toaddress privacy concerns around linkability of the PvD HTTP connectionwith future user-initiated connections, if the host has a temporary addressper[RFC4941] in this PvD, then itSHOULD use a temporary addressto fetch the PvD Additional Information andMAY deprecate theusedtemporary address and generate a new temporary address afterward.¶
If the HTTP status of the answer is greater than or equal to 400, the hostMUST close its connection and consider that there is no PvD Additional Information. If the HTTP status of the answer is between 300 and 399, inclusive, itMUST follow the redirection(s). If the HTTP status of the answer is between 200 and 299, inclusive, the response is expected to be a single JSON object.¶
After retrieval of the PvD Additional Information, hostsMUST rememberthe last Sequence Number value received in an RA including the samePvD ID. Whenever a new RA for the same PvD is received with a differentSequence Number value, or whenever the expiry date for the additionalinformation is reached, hostsMUST deprecate the AdditionalInformationand stop using it.¶
Hosts retrieving a new PvD Additional Information objectMUST checkfor the presence and validity of the mandatory fields specified inSection 4.3. A retrieved object including anexpirationtime that is already past or missing a mandatory elementMUSTbeignored.¶
In order to avoid synchronized queries toward the server hostingthe PvD Additional Information when an object expires, object updatesare delayed by a randomized backoff time.¶
In the example inFigure 2, the Delay field value is 1; this means that the host calculates its delay by choosing a uniformly random time between 0 and 2(10 + 1) milliseconds, i.e., between 0 and 2048 milliseconds.¶
Since the Delay value is directly within the PvD Option rather than the object itself, an operator may perform a push-based update by incrementing the Sequence Number value while changing the Delay value depending on the criticality of the update and the capacity of its PvD Additional Information servers.¶
In addition to adding a random delay when fetching Additional Information, hostsMUST enforce a minimum time between requesting AdditionalInformationfor a given PvD on the same network. This minimum time isRECOMMENDEDto be 10 seconds, in order to avoid hosts causing a denial-of-service on thePvD server. Hosts alsoMUST limit the number of requests thatare made todifferent PvD Additional Information servers on the same network within ashortperiod of time. ARECOMMENDED value is to issue no more thanfive PvDAdditional Information requests in total on a given network within 10 seconds.For more discussion, seeSection 6.¶
The PvD Additional Information object includes a set of IPv6prefixes (under the key "prefixes") thatMUST be checkedagainst allthe Prefix Information Options advertised in the RA. If any of theprefixes included in any associated PIO is not covered by at least one of thelisted prefixes, the PvD Additional InformationMUST beconsideredto be a misconfiguration andMUST NOT be used by the host. SeeSection 4.4 for more discussion on handlingsuch misconfigurations.¶
If the request for PvD Additional Information fails due to a TLS certificate validationerror, an HTTP error, or because the retrieved file does not contain valid PvDJSON,hostsMUST close any connection used to fetch the PvDAdditional InformationandMUST NOT request the information for that PvD ID again forthe durationof the local network attachment. If a host detects 10 or more such failuresto fetch PvD Additional Information, the local network is assumed to bemisconfigured or under attack and the hostMUST NOT make anyfurtherrequests for any PvD Additional Information, belonging to any PvD ID, forthe duration of the local network attachment. For more discussion, seeSection 6.¶
Whenever the H-flag is set in the PvD Option, a valid PvD Additional Information objectMUST be made available to all hosts receiving the RA by the network operator. In particular, when a captive portal is present, hostsMUST still be allowed to perform DNS, certificate validation, and HTTP-over-TLS operations related to the retrieval of the object, even before logging into the captive portal.¶
RoutersSHOULD increment the PvD Option Sequence Number by one whenever a new PvD Additional Information object is available and should be retrieved by hosts. If the value exceeds what can be stored in the Sequence Number field, itMUST wrap back to zero.¶
The server providing the JSON filesSHOULD also check whether the client address is contained by the prefixes listed in the Additional Information andSHOULD return a 403 response code if there is no match.¶
The PvD Additional Information is a JSON object.¶
The following table presents the mandatory keys, whichMUST be included in the object:¶
| JSON key | Description | Type | Example |
|---|---|---|---|
| identifier | PvD ID FQDN | String | "pvd.example.com." |
| expires | Date after which this object is no longer valid | [RFC3339] Date | "2020-05-23T06:00:00Z" |
| prefixes | Array of IPv6 prefixes valid for this PvD | Array of strings | ["2001:db8:1::/48", "2001:db8:4::/48"] |
A retrieved object that does not include all three of these keys atthe root of the JSON objectMUST be ignored. All three keysneedto be validated; otherwise, the objectMUST be ignored. Thevalue storedfor "identifier"MUST be matched against the PvD ID FQDNpresented in thePvD Option using the comparison mechanism described inSection 3.4.The value stored for "expires"MUST be a valid date in thefuture.If the PIO of the received RA is not covered by at least one of the "prefixes"key, the retrieved objectSHOULD be ignored.¶
The following table presents some optional keys thatMAY beincluded in the object.¶
| JSON key | Description | Type | Example |
|---|---|---|---|
| dnsZones | DNS zones searchable and accessible | Array of strings | ["example.com", "sub.example.com"] |
| noInternet | No Internet; set to "true" when the PvD is restricted | Boolean | true |
It is worth noting that the JSON format allows for extensions.Whenever an unknown key is encountered, itMUST be ignoredalong withits associated elements.¶
Private-use or experimental keysMAY be used in the JSONdictionary. In order to avoid such keys colliding with the keys registered byIANA,implementers or vendors defining private-use or experimentalkeysMUST create sub-dictionaries. If a set of PvD AdditionalInformation keysare defined by an organization that has a formal URN namespace[IANA-URN],the URN namespaceSHOULD be used as the top-level JSON key forthe sub-dictionary. For other private uses, the sub-dictionary keySHOULD follow the format of "vendor-*", where the "*" isreplaced by theimplementer's or vendor's identifier. For example, keys specific to the FooBarorganization could use "vendor-foobar". If a host receives a sub-dictionarywithan unknown key, the hostMUST ignore the contents of the sub-dictionary.¶
The following two examples show how the JSON keys defined in thisdocument can be used:¶
{ "identifier": "cafe.example.com.", "expires": "2020-05-23T06:00:00Z", "prefixes": ["2001:db8:1::/48", "2001:db8:4::/48"],}{ "identifier": "company.foo.example.com.", "expires": "2020-05-23T06:00:00Z", "prefixes": ["2001:db8:1::/48", "2001:db8:4::/48"], "vendor-foo": { "private-key": "private-value", },}¶HostsMUST validate the TLS server certificate when retrieving PvDAdditional Information, as detailed inSection 4.1.¶
HostsMUST verify that all prefixes in all the RA PIOs are covered by aprefix from the PvD Additional Information. An adversarial routerattempting to spoof the definition of an Explicit PvD, without the ability tomodify the PvD Additional Information, would need to perform IPv6-to-IPv6NetworkPrefix Translation (NPTv6)[RFC6296] in orderto circumvent this check.Thus, this check cannot prevent all spoofing, but it can detectmisconfigurationor mismatched routers that are not adding a NAT.¶
If NPTv6 is being added in order to spoof PvD ownership, the HTTPS server for Additional Information can detect this misconfiguration. The HTTPS serverSHOULD validate the source addresses of incoming connections (seeSection 4.1). This check gives reasonable assurance that NPTv6 was not used and restricts the information to the valid network users.If the PvD does not provision IPv4 (it does not include the L-flag in the RA), the server cannot validate the source addresses of connections using IPv4. Thus, the PvD ID FQDN for such PvDsSHOULD NOT have a DNS A record.¶
This section describes some example use cases of PvDs. For the sake ofsimplicity, the RA messages will not be described in the usual ASCII artbut rather in an indented list. Values in the PvD Option header that are notincluded in the example are assumed to be zero or false (such as theH-flag, Sequence Number, and Delay fields).¶
In this example, there is one RA message sent by the router. This message contains some options applicable to all hosts on the network and also a PvD Option that also contains other options only visible to PvD-aware hosts.¶
PvD Option header: length = 3 + 5 + 4, PvD ID FQDN =example.org., R-flag = 0 (actual length of the header with padding24 bytes = 3 * 8 bytes)¶
Note that a PvD-aware host will receive two different prefixes,2001:db8:cafe::/64 and2001:db8:f00d::/64, both associated with the same PvD (identified by "example.org."). A non-PvD-aware host will only receive one prefix,2001:db8:cafe::/64.¶
It is expected that for some years, networks will have a mixedenvironment of PvD-aware hosts and non-PvD-aware hosts. If there is aneed to give specific information to PvD-aware hosts only, then it isRECOMMENDED to send two RA messages, one for each class ofhosts.This approach allows for two distinct sets of configuration informationto be sent in a way that will not disrupt non-PvD-aware hosts. It alsolowers the risk that a single RA message will approach its MTU limit dueto duplicated information.¶
If two RA messages are sent for this reason, theyMUST be sent from twodifferent link-local source addresses (Section 3.2). For example, here is theRA sent for non-PvD-aware hosts:¶
PvD Option header: length = 3 + 2, PvD ID FQDN = foo.example.org., R-flag = 1 (actual length of the header 24 bytes = 3 * 8 bytes)¶
And here is the RA sent for PvD-aware hosts:¶
PvD Option header: length = 3 + 2 + 4 + 3, PvD ID FQDN = bar.example.org., R-flag = 1 (actual length of the header 24 bytes = 3 * 8 bytes)¶
In the above example, non-PvD-aware hosts will only use the first listed RA sent by their default router and use the2001:db8:cafe::/64 prefix. PvD-aware hosts will autonomously configure addresses from both PIOs but will only use the source address in2001:db8:f00d::/64 to communicate past the first-hop router since only the router sending the second RA will be used as the default router; similarly, they will use the DNS server2001:db8:f00d::53 when communicating from this address.¶
In this example, the goal is to have one prefix from one RA be usable by both non-PvD-aware and PvD-aware hosts and to have another prefix usable only by PvD-aware hosts. This allows PvD-aware hosts to be able to effectively multihome on the network.¶
The first RA is usable by all hosts. The only difference for PvD-aware hosts is that they can explicitly identify the PvD ID associated with the RA. PvD-aware hosts will also use this prefix to communicate with non-PvD-aware hosts on the same network.¶
The second RA contains a prefix usable only by PvD-aware hosts. Non-PvD-awarehosts will ignore this RA; hence, only the PvD-aware hosts will be multihomed.¶
PvD Option header: length = 3 + 2 + 4 + 3, PvD ID FQDN = bar.example.org., R-flag = 1 (actual length of the header 24 bytes = 3 * 8 bytes)¶
Note: the above examples assume that the router has received its PvD IDs from upstream routersor via some other configuration mechanism. Another document could define waysfor the routerto generate its own PvD IDs to allow the above scenario in the absence of PvDID provisioning.¶
In this example, the router indicates that it provides Additional Information using the H-flag.The Sequence Number on the PvD Option is set to 7 in this example.¶
A PvD-aware host will fetch <https://cafe.example.com/.well-known/pvd> to get the additional information. The following example shows a GET request that the host sends, in HTTP/2 syntax[RFC7540]:¶
:method = GET:scheme = https:authority = cafe.example.com:path = /.well-known/pvdaccept = application/pvd+json¶
The HTTP server will respond with the JSON Additional Information:¶
:status = 200content-type = application/pvd+jsoncontent-length = 116{ "identifier": "cafe.example.com.", "expires": "2020-05-23T06:00:00Z", "prefixes": ["2001:db8:cafe::/48"],}¶At this point, the host has the PvD Additional Information and knows the expiry time. When either the expiry time passes or a new Sequence Number is provided in an RA, the host will re-fetch the Additional Information.¶
For example, if the router sends a new RA with the Sequence Number set to 8,the host will re-fetch the Additional Information:¶
However, if the router sends a new RA, but the Sequence Number has not changed,the host would not re-fetch the Additional Information (until and unless theexpiry timeof the Additional Information has passed).¶
Since the PvD Option can contain an RA header and other RA options,any security considerations that apply for specific RA options continue toapply when used within a PvD Option.¶
Although some solutions such as IPsec or SEcure Neighbor Discovery (SeND)[RFC3971] can be used in order to secure the IPv6 Neighbor Discovery Protocol, in practice, actual deployments largely rely on link-layer or physical-layer security mechanisms (e.g., 802.1x[IEEE8021X]) in conjunction with RA-Guard[RFC6105].¶
If multiple RAs are sent for a single PvD to avoid fragmentation, dropping packetscan lead to processing only part of a PvD Option, which could lead to hostsreceiving only part of the contained options. As discussed inSection 3.2, routersMUST include the PvD Option in all fragments generated.¶
This specification does not improve the Neighbor Discovery Protocolsecurity model but simply validates that the owner of the PvD FQDNauthorizes its use with the prefix advertised by the router. Incombination with implicit trust in the local router (if present), thisgives the host some level of assurance that the PvD is authorized foruse in this environment. However, when the local router cannot betrusted, no such guarantee is available.¶
It must be noted thatSection 4.4 of this documentonly provides reasonable assurance against misconfiguration but does notprevent a hostile network access provider from advertising incorrectinformation that could lead applications or hosts to select a hostile PvD.However, a host that correctly implements the multiple PvD architecture[RFC7556]using the mechanism described in this document will be less susceptible tosome attacks than a host that does not by being able to check for the variousmisconfigurations or inconsistencies described in this document.¶
Since expiration times provided in PvD Additional Information use absolute time, these values can be skewed due to clock skew or for hosts without an accurate time base. Such time valuesMUST NOT be used for security-sensitive functionality or decisions.¶
An attacker generating RAs on a local network can use the H-flag and the PvD IDto cause hosts on the network to make requests for PvD Additional Informationfrom servers. This can become a denial-of-service attack, in which an attackercan amplify its attack by triggering TLS connections to arbitrary servers inresponseto sending UDP packets containing RA messages. To mitigate this attack, hostsMUST:¶
Details are provided inSection 4.1. This attack can be targeted at generic web servers,in which case the host behavior of stopping requesting for any server thatdoesn'tbehave like a PvD Additional Information server is critical. Limiting requestsfora specific PvD ID might not be sufficient if the attacker changes the PvD IDvaluesquickly, so hosts also need to stop requesting if they detect consistentfailure whenon a network that is under attack. For cases in which an attacker is pointinghosts ata valid PvD Additional Information server (but one that is not actuallyassociatedwith the local network), the serverSHOULD reject any requeststhat do not originatefrom the expected IPv6 prefix as described inSection 4.2.¶
Retrieval of the PvD Additional Information over HTTPS requires early communications between the connecting host and a server that may be located further than the first-hop router. Although this server is likely to be located within the same administrative domain as the default router, this property can't be ensured. To minimize the leakage of identity information while retrieving the PvD Additional Information, hostsSHOULD make use of an IPv6 temporary address andSHOULD NOT include any privacy-sensitive data, such as a User-Agent header field or an HTTP cookie.¶
Hosts might not always fetch PvD Additional Information, depending on whether or not they expect to use the information. However, if a host allows requesting Additional Information for certain PvD IDs, an attacker could send various PvD IDs in RAs to detect which PvD IDs are allowed by the client. To avoid this, hostsSHOULD either fetch Additional Information for all eligible PvD IDs on a given local network or fetch the information for none of them.¶
From a user privacy perspective, retrieving the PvD Additional Informationis not different from establishing a first connection to a remoteserver or even performing a single DNS lookup. For example, mostoperating systems already perform early queries to static web sites,such as <http://captive.example.com/hotspot-detect.html>, in order todetect the presence of a captive portal.¶
The DNS queries associated with the PvD Additional InformationMUSTuse the DNS servers indicated by the associated PvD, as described inSection 4.1. This ensures the name of the PvDAdditional Information serveris not unintentionally sent on another network, thus leaking identifyinginformation about the networks with which the client is associated.¶
There may be some cases where hosts, for privacy reasons, shouldrefrain from accessing servers that are located outside a certainnetwork boundary. In practice, this could be implemented as an allowed listof 'trusted' FQDNs and/or IP prefixes that the host is allowed tocommunicate with. In such scenarios, the hostSHOULD check thattheprovided PvD ID, as well as the IP address that it resolves into, arepart of the allowed list.¶
Network operatorsSHOULD restrict access to PvD AdditionalInformation to only expose it to hosts that are connected to the localnetwork, especially if the Additional Information would provide informationabout local network configuration to attackers. This can be implemented byallowing access from the addresses and prefixes that the router providesfor the PvD, which will match the prefixes contained in the PvD AdditionalInformation. This technique is described inSection 4.2.¶
IANA has removed the'reclaimable' tag for value 21 for the PvD Option in the "IPv6 Neighbor Discovery Option Formats" registry.¶
IANA has added a new entry in the "Well-Known URIs" registry[RFC8615] with the following information:¶
URI suffix: pvd¶
Change controller: IETF¶
Specification document: RFC 8801¶
Status: permanent¶
Related information: N/A¶
IANA has created and will maintain a new registry called "Additional Information PvD Keys", which reserves JSON keys for use in PvD Additional Information. The initial contents of this registry are given inSection 4.3 (both the table of mandatory keys and the table of optional keys).¶
The status of a key as mandatory or optional is intentionally not denoted in the table to allow for flexibility in future use cases. Any new assignments of keys will be considered as optional for the purpose of the mechanism described in this document.¶
New assignments in the "Additional Information PvD Keys" registry will be administered by IANA through Expert Review[RFC8126]. Experts are requested to ensure that defined keys do not overlap in names or semantics and that they represent non-vendor-specific use cases. Vendor-specific keysSHOULD use sub-dictionaries, as described inSection 4.3.¶
IANA has placed the "Additional Information PvD Keys" registry within a new registry entitled "Provisioning Domains (PvDs)".¶
IANA has also created and will maintain a new registry entitled "PvD Option Flags". This new registry reserves bit positions from 0 to 11 to be used in the PvD Option bitmask. This document assigns bit positions 0, 1, and 2 as shown in the table below. Future assignments require Standards Action[RFC8126].¶
| Bit | Name | Reference |
|---|---|---|
| 0 | H-flag | RFC 8801 |
| 1 | L-flag | RFC 8801 |
| 2 | R-flag | RFC 8801 |
| 3-11 | Unassigned |
Since these flags apply to an IPv6 Router Advertisement Option, IANA has placed this registry under the existing "Internet Control Message Protocol version 6 (ICMPv6) Parameters" registry and provided a link on the new "Provisioning Domains (PvDs)" registry.¶
This document registers the media type for PvD JSON text,"application/pvd+json".¶
Many thanks toMarkus Stenberg andSteven Barth for their earlier work on[MPVD-DNS], as well as toBasile Bruneau, who was author of an early draft version of this document.¶
Thanks also toMarcus Keane,Mikael Abrahamsson,Ray Bellis,Zhen Cao,Tim Chown,Lorenzo Colitti,Michael Di Bartolomeo,Ian Farrer,Phillip Hallam-Baker,Bob Hinden,Tatuya Jinmei,Erik Kline,Ted Lemon,Paul Hoffman,Dave Thaler,Suresh Krishnan,Gorry Fairhurst,Jen Lenkova,Veronika McKillop,Mark Townsley, andJames Woodyatt for useful and interesting discussions and reviews.¶
Finally, special thanks toThierry Danis for his valuable input and implementation efforts,Tom Jones for his integration effort into the NEAT project, andRigil Salim for his implementation work.¶