]> Apple Inc.

One Apple Park Way Cupertino California 95014 United States of America mellon@fugue.com

IoTconsultancy.nl

Utrecht Netherlands esko.dijk@iotconsultancy.nl

int dnssd mDNS TSR conflict resolution EDNS DNS This document specifies a new conflict resolution mechanism for Multicast DNS (mDNS), for use in cases where the advertisement is being proxied, rather than advertised directly, e.g. when using a combined DNS-SD advertising proxy and SRP registrar. A new EDNS option Time Since Received (TSR) is defined that communicates the time at which the set of resource records on a particular DNS owner name was most recently updated. About This Document The latest revision of this draft can be found at . Status information for this document may be found at . Discussion of this document takes place on the WG Working Group mailing list (), which is archived at . Subscribe at . Source for this draft and an issue tracker can be found at .

Introduction Unlike the Domain Name System , with its authority servers and delegation of authority, Multicast DNS has no single source of authority. Because of this, mDNS has a mechanism, conflict resolution () for detecting and fixing conflicts in mDNS advertisements. The current goal of mDNS conflict resolution is to prevent a newly advertised service from taking the place of an existing service with the same name that is already being advertised. This goal, however, assumes that the entity advertising an mDNS service is in fact authoritative for that service. For mDNS proxies, such as an advertising proxy , this is not the case: the source of truth for the service being advertised is a DNSSD Service Registration Protocol (SRP) requester. On a link with more than one SRP registrar, an SRP requester may register with one SRP registrar, and then subsequently update its registration on a different SRP registrar. Both SRP registrars may be acting as advertising proxies. If so, the original SRP registrar may still be advertising the old SRP registration using mDNS. If the information in the new SRP registration is identical to that in the old registration, this is often not a problem. However if some information has changed (e.g., a new IP address has been added, or a TXT record updated), then the new registration will be seen to be in conflict with the old registration. In addition, the method used in mDNS to detect conflicts can sometimes produce apparent conflicts where no actual conflict exists because of the way records in mDNS packets are marshalled. In the case of such an apparent conflict, the current behavior of mDNS is for the older (stale) registration to win, and the newer (current) information to be discarded. This behavior, which is entirely correct for services that are advertising on their own behalf, is exactly wrong when a service advertisement is being proxied.

Current Behavior When a new service is to be advertised, the registrant (the entity requesting the registration) typically registers the service with a central mDNS registrar on the host on which it is running. This mDNS registrar locally stores services that have already been registered, and may detect a conflict with one of those registered services. This can be true whether the conflicting service entry is data for which the mDNS registrar is authoritative (stored in its local registration database), or an entry it has received via mDNS (stored in its cache). In the case of such a conflict, no network transaction is required: the mDNS registrar detects it locally. It addresses the conflict in one of two ways. The first alternative is that the mDNS registrar will report the conflict to the registrant as an error, which it must fix. Alternatively, if the registrant has indicated that the mDNS registrar should automatically choose a new name for the service in case of conflict, the mDNS registrar does so automatically, without necessarily notifying the registrant. Once any locally-detectable conflicts have been resolved, the mDNS registrar probes (see ) the local network to see if any other host has already registered a service name that conflicts with the proposed new service name. If such a service name is present on the network, the mDNS registrar follows the same process previously described: either report the error to the mDNS registrant or automatically choose a new name. The effect of this approach is that generally whichever registrant first registers a service under a particular name wins. If another registrant comes along later and registers the same service name with conflicting service information, the newcomer’s information is rejected.

Problem Statement The current behavior works well for registrants registering services on their own behalf. However, for example in the case of an SRP registrar, it works poorly: an SRP registrar acting as an advertising proxy publishes the contents of its DNS zone using mDNS. The sources of truth for this information are the SRP requesters, not the SRP registrar itself. The SRP registrar in this case acts as a proxy for the SRP requesters. In the case of an advertising proxy publishing records originating from an SRP Update, the most recent information is correct; the older information is simply stale, and not competing. When the SRP requester is able to continue registering with the same SRP registrar, this works well: stale data is automatically removed and replaced with current data. However, if more than one SRP registrar is available, the SRP requester may wind up sending its SRP Updates to a different SRP registrar. This can happen as a result of a network partition event, or in cases where the SRP registrar is contacted using an anycast address. When the SRP requester sends its SRP Update to a different SRP registrar, the behavior of the mDNS conflict resolution approach without TSR is that the SRP requester's service will be given a new name, and both the old (stale) service advertisement (A) and the new (more recent) service advertisement (A’) will be discoverable as separate services. This creates a new burden on consumers of such services: they need to parse through the whole list of services of their type, using metadata from the TXT record in the service instance data, if possible, to determine that service A and service A’ are the same service. If no such information is present in the TXT record, the only way to determine that one of these two advertised services is stale is to attempt to use the advertised service, which may no longer be reachable if, for example, the change that produced the conflict was an IP address change. When the SRP lease for the stale service expires, that service's mDNS advertisement will be removed, and the service will no longer be discoverable under the original name, even if its IP address hasn't changed. This document proposes an enhancement to the current conflict resolution algorithm for mDNS, which allows an mDNS proxy to report the time at which it received the registration for DNS records it is advertising, and the source from which these records were received. This is done using a new Time Since Received (TSR) EDNS(0) () option, of which there must be exactly one per name being advertised by the mDNS proxy.

Conventions, Terms and Definitions 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 when, and only when, they appear in all capitals, as shown here. The list below defines the most important terms that are used in this document. In addition, a reader is expected to be familiar with DNS terminology , in particular "EDNS", "OPT", "Owner", "resource record", "RR", and "RRset".

absolute time

a node-local timestamp that is derived from a local monotonic clock. This clock is not required to be synchronized with any other clock, nor is it an absolute time in the sense of UTC. The clock may restart upon boot of the node but not during regular operation.

authoritative data

DNS records stored in the local registration database, for which an mDNS registrar is authoritative (as defined by ).

cache

each mDNS registrar maintains a cache of DNS records received from hosts on the local network, as described in . Information in the cache represents the mDNS registrar's current understanding of what records are advertised on the network. Authoritative data may be present in the cache, but the presence of data in the cache does not in itself indicate that the mDNS registrar is authoritative for that data.

local registration database

a database of DNS records maintained by the mDNS registrar. Records in this database are registered locally with the mDNS registrar. The mDNS registrar is authoritative for all DNS records in this database. When the mDNS registrar receives queries that match records in this database, the mDNS registrar can respond to these queries with the matching records.

mDNS registrar

an mDNS implementation on a host that accepts local requests for querying, advertising and registering DNS records from one or more requesters and/or registrants. This could for example be an mDNS daemon process running in an operating system, accepting API calls from local processes to register or update DNS records for that process. It stores the locally-registered records in its local registration database.

mDNS registrant

an entity or software process that is attempting to register records for advertisement to a (local) mDNS registrar.

mDNS requester

an entity or software process that issues mDNS queries to a local mDNS registrar, via an API call. The mDNS registrar is responsible for executing these queries and notifying the mDNS requester about the answer(s).

mDNS proxy

a host that runs an mDNS registrar and at least one mDNS registrant acting as a proxy. That is, it needs to advertise mDNS records on behalf of one or more entities not located on the host itself. The advertising proxy is an example of an mDNS proxy.

SRP requester

a client that uses the Service Registration Protocol (SRP) to send an SRP Update to an SRP registrar.

SRP registrar

a server that accepts SRP Updates sent by SRP requesters using the SRP . DNS records registered via SRP to an SRP registrar may then be advertised by mDNS using an advertising proxy located on the same host. In that case, the SRP registrar process acts as an mDNS registrant towards its local mDNS registrar process.

TSR data

locally stored data, associated with a single DNS owner name, that keeps track of the absolute time ('TSR time') when a set of resource records was last updated and includes a key checksum to identify the owner of these records. It may be data proposed by an mDNS registrant's registration, or stored in the local registration database, or extracted from a received mDNS advertisement by parsing a TSR option.

TSR option

an EDNS0 Time Since Received (TSR) option as defined by this specification.

TSR time

the (node-local) absolute time indicated by a particular TSR data. The internal representation of TSR time is defined in , and comparisons in .

Time Since Received EDNS Option Each Time Since Received (TSR) EDNS option is applicable to exactly one DNS owner name. So all the records for that owner name that appear in the answer, authority and/or additional sections of an mDNS message would be covered by a single TSR option. The TSR EDNS option has the following format: The TSR EDNS option includes three fields in the OPTION-DATA (): 'RR Index' (two-byte integer in network byte order), 'Key Checksum' (four bytes), and 'Time Offset' (four bytes). The 'RR Index' is the number of the RR in the mDNS packet. Question RRs are not counted. So if the message includes two answer RRs, one authority RR and two additional RRs, an index of 0 would refer to the first answer, an index of 1 to the second answer, an index of 2 to the single authority record, and so on. Questions are excluded because they have no data associated with them, and so it makes no sense to have TSR options associated with them. If there is more than one record in the mDNS message with the same owner name, only one TSR option is emitted for that name, and it applies to every RR in the mDNS message with that owner name. It is not possible in the SRP protocol for two updates at two different times to contain records that apply to the same name: in such a situation, the second update completely replaces the first, so all data in the first update is then rendered stale. The second field, 'Key Checksum', is a simple 32-bit checksum of the public key that the owner of the data (for example the SRP requester) used to authenticate itself. The key checksum is computed by treating the key as a series of 32-bit unsigned integers in network byte order, and adding these integers together to produce a 32-bit unsigned checksum. Overflow is not considered. This checksum need not be cryptographically secure: mDNS messages are not authenticated, so an attacker on the local link can always cause problems with mDNS by providing spurious responses. The purpose of the checksum is simply to notice whether, for a specific owner name, two different authoritative sources have provided information. The 'Time Offset' field contains the difference, in seconds, between the time at which the TSR option is being generated and the time of receipt of resource records for that owner name (as encoded in TSR data). The time offset is represented in the mDNS message as a time in seconds relative to the time when the mDNS message is sent. If this difference is greater than 7 days (7 * 24 * 60 * 60 = 604,800), the mDNS registrar MUST use a value of 7 days rather than the larger value. The relative time value in the TSR option is converted to an absolute (local) time when stored in a cache or a local registration database on an mDNS registrar as TSR data, and is converted back to a relative time whenever an mDNS message with a TSR option is generated from local data.

mDNS Registrar Behavior

Validating requested local RR registrations with TSR data When a local mDNS registrant asks an mDNS registrar to register one or more records on an owner name, and provides TSR data for that name, the mDNS registrar first checks that none of the records are marked shared. If any record is marked shared, the mDNS registrar MUST respond with an error indicating that the registration is invalid. It then checks to see if there are any records either in cache or in its local registration database on that owner name. If no such data exists, the mDNS registrar stores the record(s) in this registration in its local registration database and puts them in the probing state. When such data exists, the registrar MUST check to see if it has TSR data for that owner name. If it does not, or if there is TSR data on that name but the key checksum does not match, the registrar MUST treat this registration as a conflict and return an appropriate error to the registrant. If such data exists and the key checksums match, there are three possibilities based on the known TSR time (from the existing data) and the proposed TSR time (from the TSR data in the registration request):

Known TSR time is more recent

In this case, the registrar MUST treat the requested data as stale, and return an indication to the registrant that its registration is stale. This indication must be distinct from the "appropriate error" indication of conflict that was defined above.

Both TSR times are the same (equal)

In this case, cached data (if any) on the owner name is discarded first. Then, the requested records are added to the local registration database, updating existing records (if any). Since the requested records are considered identical to the known records, based on TSR time, no probing or announcing is performed for these records.

Proposed TSR time is more recent

In this case, cached data (if any) on the owner name is discarded first. The registrant for any existing locally-registered data is notified that the data they had previously registered is stale, and the stale data (if any) is removed from the local registration database. The requested records including TSR data are then added to the local registration database. The records are put in the probing state only if needed, as described in . If probing is not performed, the new records are only announced () and a "goodbye" announcement () with TSR option is sent for any RRset that is removed due to the local registration database update.

Note that any TSR time comparisons are performed with a small tolerance as defined in . It is in principle possible for two different mDNS registrants to ask the same mDNS registrar to publish different RRs on the same name, some of which are shared and some of which are unique (see ). If an mDNS registrant tries to register an RR on a name for which the mDNS registrar already has data, cached or authoritative, on the same name, whether of the same RR type or a different RR type, for which there is no TSR data, or for which the key checksum in the TSR data being registered does not match what is already known, the registrar MUST treat this as an immediate conflict, and MUST NOT add the data to its local registration database and MUST NOT probe.

Probing or Updating when the proposed TSR time is more recent For the third case above where the proposed TSR time is more recent, the newly registered records are only put in the probing state if the mDNS registrar does not have unique ownership of the name yet. This follows : a host that has already established unique ownership of a name does not repeat the probing step when it updates the records on that name, but only re-announces them. Therefore, an mDNS registrar that has already completed probing for the previous (now stale) set of records on this owner name, that were stored in the local registration database, MUST NOT probe the newly registered records on this same name. Cases where the mDNS registrar has not (yet) established unique ownership of the owner name include for example:

• previous records on the name were present only in the cache (learned from another mDNS host on the link);
• records on the name are already locally registered but still in the probing state (i.e., probing has not completed yet).

In any such cases the mDNS registrar MUST put the requested records in the probing state and treat them as tentative, as with any new local mDNS registration, until they have been fully probed as described in . If probing for the owner name was already in progress, it SHOULD continue without restarting: only the proposed records in the authority section () are updated for the probe messages that are still to be sent.

Probing resource records on names with TSR data describes how an mDNS registrar probes to ensure that there is no conflicting data for records in the probing state. The behavior for records that are in the probing state with owner names to which no TSR data applies remains unchanged. When there is TSR data on a name for which records are being probed, the mDNS registrar MUST include TSR options for each such name as described in and . Handling of responses is described in .

Probing when only the TSR time has changed If the only change to the set of mDNS records advertised on a name with TSR data is the TSR time itself, probing is not necessary and in this case an mDNS registrar MUST NOT probe. The reason to eliminate this probe is that it avoids unnecessary multicast traffic: in this situation the probe serves no purpose, since nothing substantive has changed and the mDNS registrar was already authoritative for the records.

Processing mDNS questions for which TSR data exists When processing a question for which local TSR data is present, the mDNS registrar MUST first check to see if there is corresponding data in the mDNS message being processed. In this case, before constructing a response, the mDNS registrar MUST process the non-question records in the packet, since this could result in stale data being flushed. Processing is performed as described in . Once all non-question records have been processed, the registrar MUST respond to any questions that match locally-registered resource records for which a known answer is not present in the query. Responses are constructed as described in .

Receiving mDNS messages that may contain TSR options mDNS registrars that support TSR need to compute an absolute time based on a time offset. This means that registrars need to know when the packet was received. A naive implementation might assume that the time that the packet is read off the input queue by the registrar is close enough. However, in practice it can be the case on a heavily loaded system that the time of receipt and the time of processing are far enough apart to create the appearance of staleness. To avoid this, mDNS registrars that have an API available to get the actual time of receipt of a packet should make use of that API. For example, the SO_TIMESTAMP_CONTINUOUS socket option is available on Linux and BSD Unix platforms, including MacOS. When such APIs are not available, another option is to receive such packets on a high priority thread, mark a timestamp and queue them for later processing.

Processing mDNS messages that may contain TSR options mDNS registrars that support the TSR option MUST check incoming messages for the presence of an EDNS(0) option containing TSR options. mDNS registrars that do not support TSR will not do this check, and will behave as if no TSR options are present. For non-proxy use cases, this should make no difference, since in such cases if multiple devices advertise records on the same owner name, these are actually in conflict. However, for the proxy use case, what this means is that two proxies that are proxying the same data cannot interoperate if one supports TSR and the other doesn't. It is important to note that mDNS messages, particularly in the case of proxies, can contain combined information answering multiple queries that may be outstanding, and as such, it's entirely possible for an mDNS message sent by an mDNS registrar that supports TSR to contain some answers for which there is TSR data, and some answers for which there is not. It's equally possible that such a registrar will send mDNS packets containing no TSR options at all. When an mDNS message contains TSR options, for each TSR option in an mDNS message, the mDNS registrar first determines the owner name of the TSR option by assigning an index to each non-question resource record in the mDNS message. The 'RR Index' of each TSR option is then matched to the index of a resource record, and the owner name for that resource record is applied to the TSR option. The TSR time is then computed by taking the current local clock time and subtracting from it the 'Time Offset' value in the TSR option. If there is a TSR option in an mDNS message for which there is no matching resource record in the mDNS message, the mDNS registrar MUST ignore that TSR option. The mDNS registrar MUST NOT use the 'RR Index' value in the TSR option to search across the mDNS packet since such an index can easily be out of bounds. Now, for each record in the mDNS message, the mDNS registrar first determines whether the record is an OPT record, is in the question section, or is a known answer (QR bit is '0' and it's a record in the answer section). For all such records, no special processing is done for TSRs, since no TSR should exist in the mDNS message for these records. For each remaining resource record in the mDNS message, the mDNS registrar MUST check to see if there is a TSR option in the mDNS message for that owner name. If there is not, the mDNS registrar MUST check to see if there is TSR data with that owner name locally. If there is not, the record is processed normally. If there is local TSR data for the record's owner name, but no TSR data for that owner name in the mDNS message, the mDNS registrar checks to see if there are any resource records in the local registration database on that name. If there are, all such records are treated as in conflict. This conflict exists even if the locally registered records are all shared records. In cases where there are records on the name in the cache, those records are all discarded, because they are in conflict with the new data. In the case that there is TSR data for the record in the mDNS message, and there are local records on the same owner name for which there is no local TSR data, this always means that any data is in conflict. How that conflict is addressed depends on the data. First, note that resource records in the answer section of an mDNS query (QR bit is '0') are "known answers" and therefore are not relevant when adding data to the cache. Such records can never have TSR options associated with them. However, resource records in the authority and additional sections of a query do need to be processed (but in the case of authority records, are not added to the cache). In cases where the TSR data for a particular name is present both locally and in the mDNS message, the mDNS registrar MUST compare the key checksums. If these are different, then the records are always in conflict, and are handled according to the context of the conflict, as described in . In cases where the key checksums match, the mDNS registrar MUST compare the times. When the TSR time from the mDNS message is more recent than the local TSR time, the local data in the cache is flushed. Stale data in the local registration database is removed, and the mDNS registrant is informed that this data is stale. When the TSR times are the same, any resource records on that name in the answer section and additional section are added to the cache. When the local TSR time is more recent, the data in the message is not added to the cache, and no action is taken with respect to any locally-registered data.

Duplicate answer suppression behavior , Duplicate Answer Suppression, describes behavior intended to prevent the redundant transmission of duplicate answers. When an mDNS query is received for which the mDNS registrar has authoritative data, the mDNS registrar will wait for a random amount of time before sending a response. If, during that time, a response is received that contains all the answers it would have sent, it suppresses sending these answers, since they are redundant. Such a received response is referred to as a "pre-empting response" from here on. In the case where TSR data is present locally or in a pre-empting response, it can be the case that the data in a pre-empting response is stale or conflicting. For this reason, an mDNS registrar MUST NOT suppress duplicate answers when:

• the TSR key checksum for the owner name in a pre-empting response does not match the local TSR key checksum for that owner name
• TSR data is present locally for that owner name but is not present in the pre-empting response
• A TSR option is present for that owner name in the pre-empting response, but local authoritative data for that owner name is present with no associated TSR data
• local authoritative data for that owner name is present and includes TSR data, and a TSR option for that owner name is present in the pre-empting response, but the local TSR time for that owner name is more recent than the TSR time for that owner name in the pre-empting response

shows what this might look like from the perspective of an mDNS requester.

Suppression of "Goodbye" announcements When authoritative data is removed on an mDNS registrar because an mDNS message has been received with more recent data, the mDNS registrar MUST NOT send a "goodbye" announcement for any RR/RRset on that owner name as a result of flushing this stale data. The reason for this is that in the case where an mDNS registrant updates one or more RRsets on an owner name covered by TSR data, and as a result of this some records are removed, but some remain, the mDNS registrar that was directly updated will either send a "goodbye" announcement or an announcement with the cache flush bit set as specified in . Since the mDNS registrar with the most current information has already done what is needed, if an mDNS registrar that is flushing locally-registered data were to send a "goodbye" announcement this would at best be redundant and hence wasteful use of multicast, and at worst might cause valid data to be flushed from the cache on some mDNS registrar.

Suppression of redundant probing There are use cases for TSR in which the same set of records, with the same TSR time, can be registered at multiple mDNS registrars on multiple hosts. Examples are: * an mDNS registrant that registers its set of records to a remote SRP registrar via unicast, while at the same time it advertises these records itself, if needed, using mDNS as a backup solution. * a group of two or more mDNS proxies that run a replication protocol, each one sharing any change to a set of records with its peers. Each proxy's mDNS registrar then can advertise the same records, for redundancy. If in such cases the registration triggers probing by an mDNS registrar, the situation of multiple mDNS registrars probing at the same time for the same data needs to be avoided. Therefore, the following requirement is imposed on any mDNS registrant using TSR: the mDNS registrant (out of multiple) that first registers a set of records to an mDNS registrar MUST wait until the records are successfully registered at this mDNS registrar, before it can trigger a registration at another mDNS registrar. This makes it likely that any probing, if needed, is performed by the first mDNS registrar only and not by others. For example, if multiple mDNS registrants are peers in a replication protocol, then the replication of records from the first (origin) peer to a second peer can only be confirmed to the second peer once the local registration to the mDNS registrar on the first peer's host returns 'success' which implies the probing was done and the mDNS announce message for the new records is either already sent on the link, or is pending transmission (i.e. queued for Layer 2 multicast). It is perfectly fine if the first peer sends the record data to the second peer earlier than that (for example, while probing is ongoing) but the second peer's mDNS registrant must not register this data yet to its local mDNS registrar before the first peer confirms its local registration was successful. Due to the above requirement, it will often be the case that one or more other mDNS registrars on the same link already have the replicated records in their cache at the time that they receive the registration request from a local mDNS registrant. This cached data will then also have matching (equal) TSR times to the data being registered. In such cases, these mDNS registrars apply the second case in ("Both TSR times are the same") and no probing or announcing is performed for the data by these mDNS registrars. If an update by an mDNS registrant causes some RRsets to be removed on the mDNS registrar, then a goodbye announcement is sent for those RRsets, but no probe is sent for RRs that are not removed. If some new RRs are added in an update on a name the mDNS registrar already owns, probing can be skipped because the mDNS registrar already knows it has ownership of the name. So the new RRs can simply be announced immediately. Note that if the TSR time for a set of records is updated by an mDNS registrant, but nothing else changes, the mDNS registrar does not re-probe those records as defined in .

Handling multi-packet mDNS advertisements It can happen, particularly when some sort of synchronization event occurs, such as the announcement of a new IPv6 prefix that triggers many SRP registrations, that the mDNS records on a particular name do not all fit in the remaining space in an mDNS packet during packet construction (see ). The mDNS protocol does not require that such records all be sent in the same mDNS message. With mDNS probes and announcements that do not use TSR, this can result in conflicts when the information being announced or probed is advertised by more than one mDNS proxy. In the case of mDNS proxies using TSR, if the TSR key checksums are the same, there is no possibility of a conflict, and so this processing, including the simultaneous probe tiebreaking of and the conflict detection of , need not be performed. The TSR stale data processing defined by this document does not rely on all the records being present in a single mDNS message, and so the stale data evaluations described in previous subsections are all that is needed to handle this case. Because of this, when an mDNS registrar is processing received mDNS records on a name with TSR data where the local and received key checksums are identical, it MUST NOT treat a partial set of records in the received mDNS message as a conflict or an indication of staleness. Only the TSR time can be an indication of staleness in this case.

Constructing an mDNS message with TSR options For each non-question record that is added to the mDNS message, one of three things must be true:

• The mDNS registrar has locally-registered resource record(s) on that owner name, which may or may not be in the probing state.
• It is sending an answer which is either an announcement or a response containing data it has already validated and for which it is authoritative.
• The message is a query (QR bit is '0') and the record is in the answer section, and is therefore a "known answer."

As described in , an mDNS registrar asking a question about one or more RRs on a particular name populates the answer section of its mDNS message with the answers it already knows, to avoid unnecessary responses. However, in this case it can't also be probing for records on the same name, because probes are only done for unique (non-shared) records. The requirements in mean that there can never be an mDNS probe message that contains known answer records on an owner name for which any RR is being probed to which TSR data applies. This means that for any particular owner name that might be represented in an mDNS packet, it must be the case either that it is not a known answer, or that it is a known answer and no other records exist in the mDNS packet with the same owner name to which a TSR option would apply. That is, one of two things must be true about the set of all records with a particular owner name being added to the mDNS packet: either a TSR option applies to all these records, or it applies to none of the records. Furthermore, either a record is a known answer from cache, or it is a locally-registered record. When constructing an mDNS message, the mDNS registrar maintains a set of names and associated TSR data. Initially this set is empty. When the registrar adds a record to the mDNS message, if that record is locally registered, and if the registrar has TSR data for that name, it first checks to see if it has already added TSR data for that name to the set. If not, then it adds a new entry to the set containing the TSR data for the owner name of the RR. The data added consists of the owner name, the index of the record being added (since it is the first), the key checksum, and the time offset. Once the mDNS registrar has finished adding resource records to the mDNS message, it adds an OPT record in the additional section. In this OPT record it adds a TSR option for every name in the set that was generated when adding resource records to the message. The time of receipt is subtracted from the current time to produce the value for the 'Time Offset' field (as specified in ). This value is clamped to a maximum of seven days (604,800 seconds) . Note that "goodbye" announcements for an owner name that has TSR data associated are also covered by the second bullet above and therefore MUST have an associated TSR option in the mDNS message.

The Effect of Network Latency on Time Computations Because TSR time computations are affected by network latency as well as rounding effects (when computing the 'Time Offset' value in a TSR option), comparisons can’t be considered accurate. It is therefore necessary to tolerate some amount of error or uncertainty when comparing two TSR times. This document defines a tolerance of 2 seconds for comparing two TSR times: given two TSR times t1, t2 to compare, t1 equals t2 iff | t1 - t2 | <= 2. In practice, however, it should generally not be the case that two mDNS registrars receive a registration from a single source with different sets of records at nearly the same time. For example, two SRP registrars acting as advertising proxy that both receive an SRP update from the same SRP requester at nearly the same time with different data would be unlikely for a well-behaved SRP requester. So it should always be the case that either there is a clear ordering to the TSR 'Time Offset' values as advertised by the two mDNS registrars, or that there is no conflict in the two sets of data. In the latter case, the above equality definition with tolerance holds true. For example, when sending SRP Updates with IP anycast, an SRP Update UDP retransmission triggered by a lost SRP Update response could be routed to a different SRP registrar. In this case both SRP registrars would receive identical data with a small time difference of at least the minimum DNS over UDP retransmission interval (2-5 seconds per ). Depending on rounding effects and network delays, the resulting difference in the TSR 'Time Offset' value advertised by the mDNS registrars could be 2 seconds, in which case the set of records would be (correctly) treated as being identical and having no conflict, or it could be 3 seconds or higher, in which case there is a clear ordering to the data with the latest registration "winning", which is also correct.

Internal Handling of TSR Data The TSR 'Time Offset' value that is sent on the wire is expressed in seconds relative to the time of receipt of the registration. In order to derive this value, the mDNS registrar must remember the (local) time at which the registration occurred. This time is recorded as an absolute time, not a relative time. We refer to this as the time of receipt or 'TSR time'. The resolution of the internal TSR time is implementation-specific. Typically it is stored in a higher resolution than seconds, for example milliseconds or units of 10 milliseconds, to mitigate rounding errors as much as possible. When constructing a TSR option, the registrar computes the difference between the current time and the TSR time, which must always be in the past. This difference, which should be a positive integer, is converted to seconds, and that unsigned value is then used to synthesize the 'Time Offset' value in the TSR option. The conversion to seconds MUST use a rounding operation (i.e. not use a 'floor' or 'ceil' operation). Note that a 'Time Offset' value can be 0.

Timeliness of Conflict Resolution If a conflict exists in which some mDNS registrant(s) store stale data, it is expected that the conflict is recent, and that it will be resolved quickly since mDNS message exchanges with TSR options cause removal of the stale data. Different hosts may be able to record shorter or longer time offsets, depending on their implementation. However, because of this expectation of recentness, mDNS registrars should never need to report a TSR option 'Time Offset' value of longer than seven days. It’s reasonable to expect that every mDNS implementation should be able to remember time intervals of at least seven days. This maximum time interval is normatively defined in .

Legacy Behavior mDNS registrars and queriers that do not support the TSR option are expected to ignore the option, so they will behave as if no TSR option was sent. This may result in such registrars temporarily caching stale data. However, in the normal course of processing, more recent data will win. In cases where it does not, the Reconfirm process which is part of already works to clear stale data: since SRP registrars are expected to implement TSR, by the time a Reconfirm is attempted, all authoritative stale data should have been cleared.

When to Use TSR There are no cases where using TSR is harmful, but in the case of individual devices advertising individual services with mDNS, it may be of little benefit. The reason for this is that when two devices both claim the same name to use for advertising services, their advertisements will be seen as in conflict whether or not TSR is present. In this case, TSR does no harm, but most likely does not help.

Use of TSR with redundant proxies Use of TSR is most beneficial in the case of redundant mDNS proxies. The reason for this is that such proxies tend to publish data that could potentially produce name conflicts as a result of updates: when one proxy is publishing an old version of the data and another proxy is publishing a new version of the data, and these data differ, this can appear as a name conflict and result in renaming. So for this use case, the use of TSR is very beneficial.

Use of TSR with multihomed devices It can also happen that a multihomed device uses SRP to register when connected to one sort of network, and uses mDNS to advertise the same service when connected to another sort of network. For example, the device may have capabilities to connect to a constrained network to reduce power use, but also to occasionally connect to a WiFi network either for backup or for bulk data transfers. In this situation we have two potentially competing mDNS advertisements. The first is that of the mDNS registrar on the multihomed device, which is directly advertising its own service using mDNS. The second is the one or more advertising proxies that are advertising services registered via SRP using mDNS. Both datasets are coming from the same source, and the advertising proxy will use TSR to identify the source. To avoid spurious conflicts, a device that sometimes registers with SRP and sometimes advertises with mDNS SHOULD use TSR when advertising with mDNS.

TSR in networks with non-compliant mDNS caches Some network infrastructure devices available commercially provide "mDNS cache" services or "mDNS proxy" services that purport to allow mDNS discovery across links that are separate at the IP layer and do not share a multicast domain. These services generally cause mDNS to become unreliable in various ways, and it would be helpful to be able to use TSR to distinguish between stale data advertised by such devices, and fresh data advertised directly from the device that is authoritative for that data. Unfortunately, for the most part these devices do not cache mDNS packets, but rather cache data advertised using mDNS. As a result, if such devices do not support TSR, we get no benefit from TSR. That said, such devices SHOULD support TSR, and if they do this might prevent some of their failure modes. What this would mean would be that any DNS RRs cached or forwarded by such devices would be accompanied by any TSR data applicable to them. When forwarded with no time delay, the TSR data could be copied verbatim (following the rules for generating packets with TSR given in ). When such RRs are cached and later resent, they would also be accompanied by their TSR data. The mDNS cache service would need to record the time at which they were received. When retransmitting such cached data, the cache service would need to adjust the time offset in the TSR option, by first subtracting the time at which the cached RRs were received from the current time, and then increasing the offset in the applicable TSR option by that amount, up to the limit of 7 days. The use of such caches and forwarders is NOT RECOMMENDED. We mention the use of TSR for such use cases because it can mitigate some of the failure modes of such caches and forwarders. However, the use of Discovery Proxy in such instances is preferred: as the size of the network covered by such caching and forwarding services increases, the amount of mDNS traffic they create increases, and this is often addressed by limiting which services are available for discovery. Discovery Proxy uses unicast DNS, and therefore does not suffer from these limitations.

Registrant API Considerations When a registrant registers a service at its mDNS registrar and requests the use of a time of receipt, the registrant MUST also specify when it received the original registration. In order to support this, the API is required not only to allow the registrant to specify that TSR conflict resolution is wanted, but also to provide a way for the registrant to specify an absolute time at which the original registration was received, and the key checksum used to identify the entity that's actually authoritative for the data. This is important, for example, in the case of SRP Replication , where an SRP registrar may receive a registration from a peer during startup synchronization. This registration will have occurred at some significant amount of time in the past, and so it would be incorrect for the mDNS registrar receiving the registration to use the time that the registrant registers the service as the time of receipt.

Removing data that is still valid In some cases, a proxy may need to stop being a proxy, but may be proxying RRs that are also being proxied by one or more other proxies. In this case, if the proxy sends a "goodbye" announcement for such RRs, they will be removed from the caches of mDNS registrars that receive such announcements. To prevent this, an mDNS registrar implementation that implements TSR MUST provide a way for an mDNS registrant to indicate that such data is being withdrawn from publication by that registrant, but is still valid. When the registrant indicates that this is the case, the mDNS registrar MUST NOT send "goodbye" announcements for such data.

Primary/Secondary indication When more than one proxy is authoritative for a particular RR, this can generate excessive answer traffic, and also redundant "goodbye" announcements. mDNS registrar implementations that support TSR MUST provide a way for a new proxy mDNS registrant to indicate that it is primary or secondary. When an RRset registered by a secondary proxy is later removed, the mDNS registrar MUST NOT send a "goodbye" announcement for that RRset. Similarly, when an RR is registered by a registrant that indicates that it is secondary, the mDNS registrar MUST NOT respond to the initial mDNS query for that RR. Only if a second mDNS query retransmission is received, within 5 seconds of the initial query reception, should it respond. Note that any change of indication from primary to secondary, or vice versa, while an mDNS registrant already has records registered on the mDNS registrar is out of scope of this specification and an API to indicate such changes is not required on the mDNS registrar.

Security Considerations The TSR option is an optimization: it ameliorates an edge case for mDNS proxies. A malicious host on the same link could use the TSR option to win conflict resolution processes. However, because TSR is only used by proxies, this technique will not work for normal mDNS service registrations: in that case, normal mDNS conflict resolution is done, and the attacker gains no benefit from using TSR. Whether or not an mDNS registration has a recorded time of receipt, an attacker can deny service by announcing its own conflicting data and then answering the subsequent probe as described in . Because it does not include a TSR option in its authority section, it can win the simultaneous conflict resolution process that follows its bogus announcement. So the TSR-based conflict resolution process creates no new vulnerability. Addressing the existing vulnerability is out of scope for this document. Protocols that rely on mDNS MUST NOT assume that mDNS service is secure or private. If security (authentication, authorization and/or secrecy) is needed, these must be provided at the application layer, or by using DNSSEC rather than mDNS for service discovery.

IANA Considerations IANA is requested to allocate a new OPT RR option code from the DNS EDNS0 Option Codes (OPT) registry, in the Domain Name System (DNS) Parameters registry group, for the 'Time Since Received' Option.

References Normative References This RFC is the revised basic definition of The Domain Name System. It obsoletes RFC-882. This memo describes the domain style names and their used for host address look up and electronic mail forwarding. It discusses the clients and servers in the domain name system and the protocol used between them. This RFC is the revised specification of the protocol and format used in the implementation of the Domain Name System. It obsoletes RFC-883. This memo documents the details of the domain name client - server communication. As networked devices become smaller, more portable, and more ubiquitous, the ability to operate with less configured infrastructure is increasingly important. In particular, the ability to look up DNS resource record data types (including, but not limited to, host names) in the absence of a conventional managed DNS server is useful. Multicast DNS (mDNS) provides the ability to perform DNS-like operations on the local link in the absence of any conventional Unicast DNS server. In addition, Multicast DNS designates a portion of the DNS namespace to be free for local use, without the need to pay any annual fee, and without the need to set up delegations or otherwise configure a conventional DNS server to answer for those names. The primary benefits of Multicast DNS names are that (i) they require little or no administration or configuration to set them up, (ii) they work when no infrastructure is present, and (iii) they work during infrastructure failures. The Domain Name System's wire protocol includes a number of fixed fields whose range has been or soon will be exhausted and does not allow requestors to advertise their capabilities to responders. This document describes backward-compatible mechanisms for allowing the protocol to grow. This document updates the Extension Mechanisms for DNS (EDNS(0)) specification (and obsoletes RFC 2671) based on feedback from deployment experience in several implementations. It also obsoletes RFC 2673 ("Binary Labels in the Domain Name System") and adds considerations on the use of extended labels in the DNS. This document specifies a network proxy that uses Multicast DNS to automatically populate the wide-area unicast Domain Name System namespace with records describing devices and services found on the local link. The Domain Name System (DNS) is defined in literally dozens of different RFCs. The terminology used by implementers and developers of DNS protocols, and by operators of DNS systems, has changed in the decades since the DNS was first defined. This document gives current definitions for many of the terms used in the DNS in a single document. This document updates RFC 2308 by clarifying the definitions of "forwarder" and "QNAME". It obsoletes RFC 8499 by adding multiple terms and clarifications. Comprehensive lists of changed and new definitions can be found in Appendices A and B. The Service Registration Protocol (SRP) for DNS-based Service Discovery (DNS-SD) uses the standard DNS Update mechanism to enable DNS-SD using only unicast packets. This makes it possible to deploy DNS-SD without multicast, which greatly improves scalability and improves performance on networks where multicast service is not an optimal choice, particularly IEEE 802.11 (Wi-Fi) and IEEE 802.15.4 networks. DNS-SD Service registration uses public keys and SIG(0) to allow services to defend their registrations. In many standards track documents several words are used to signify the requirements in the specification. These words are often capitalized. This document defines these words as they should be interpreted in IETF documents. This document specifies an Internet Best Current Practices for the Internet Community, and requests discussion and suggestions for improvements. RFC 2119 specifies common key words that may be used in protocol specifications. This document aims to reduce the ambiguity by clarifying that only UPPERCASE usage of the key words have the defined special meanings. Informative References Apple Inc. Apple Inc. IoTconsultancy.nl An Advertising Proxy advertises the contents of a DNS zone, for example maintained using the DNS-SD Service Registration Protocol (SRP), using Multicast DNS. This allows legacy clients to discover services registered with SRP using Multicast DNS. Apple Inc. Google Google This document describes a protocol that can be used for ad-hoc replication of a DNS zone by multiple servers where a single primary DNS authoritative server is not available and the use of stable storage is not desirable.

Duplicate Answer Suppression Example As described in , it is possible that two proxies may respond to an mDNS query with answers to the same question, where the data for a particular owner name in both answers is authoritative and unique, but where the TSR time in the earlier message is earlier than the TSR time in the later message. In this case, we will see the data from the earlier message added and then removed, followed by the data in the later message being added. Consider the following example using Apple's mDNSResponder implementation.

1. mDNS requester A sends a multicast query for AAAA records on the name "example.local"
2. mDNS registrar for proxy B sends a multicast response of 2001:db8:0:42::1 with TSR time T
3. mDNS registrar for proxy C sends a multicast response of 2001:db8:0:17::1 with TSR time T+300

Note that TSR times are absolute times, but these are represented in the mDNS message as relative times, so for example "TSR time T+300" when sent at time T+400 would be represented in the mDNS message as a time offset of 100 seconds, and if the message in (2) were sent at nearly the same time, it would have a time offset of about 400 seconds. In between (2) and (3) we would expect mDNS requester A to see a callback to its DNSServiceQueryRecord call, providing a AAAA record of 2001:db8:0:42::1 with the kDNSServiceFlagsAdd bit set and the kDNSServiceFlagsMoreComing bit clear. After (3) we would expect mDNS requester A to see two callbacks to its DNSServiceQueryRecord call. The first would provide a AAAA record of 2001:db8:0:42::1 with the kDNSServiceFlagsAdd bit clear and the kDNSServiceFlagsMoreComing bit set. The second would provide a AAAA record of 2001:db8:0:17::1 with the kDNSServiceFlagsAdd bit set and the kDNSServiceFlagsMoreComing bit clear. Note that we would not normally see the reverse sequence:

1. mDNS requester A sends a multicast query for AAAA records on the name "example.local"
2. mDNS registrar for proxy C sends a multicast response of 2001:db8:0:17::1 with TSR time T+300
3. mDNS registrar for proxy B sends a multicast response of 2001:db8:0:42::1 with TSR time T

This is because in this case we would expect duplicate answer suppression to result in the mDNS registrar for proxy B suppressing its response. However, it is possible that B might not see the response from C. In that situation, A would receive both responses, but since the TSR time on the second response is earlier than the TSR time on the first response, we would see only one callback, providing an AAAA record of 2001:db8:0:17::1 with the kDNSServiceFlagsAdd bit set and the kDNSServiceFlagsMoreComing bit clear.

Acknowledgments TODO acknowledge reviewers and contributors.