It is a common desire for users of XMPP to want to store their messages in a central archive on their server. This feature allows them to record conversations that take place on clients that do not support local history storage, to synchronise conversation history seamlessly between multiple clients, to read the history of a MUC room, or to view old items in a pubsub node.
As this extension aims to make things easy for client developers, some research was made into the way clients handle history today. The resulting protocol was designed to allow for the following primary usage scenarios:
Another extension for archiving already exists in XMPP, Message Archiving (XEP-0136) ). However implementation experience has shown that the protocol defined therein supports rather more functionality than is typically needed for the above uses, and is significantly more effort to implement.
This specification aims to define a much simpler and modular protocol for working with a server-side message store. Through this it is hoped to boost implementation and deployment of archiving in XMPP. It should be noted that (although not required) a server is free to implement XEP-0136 alongside this protocol if it so chooses, though a mapping between both protocols is beyond the scope of this specification.
Notable functionality in XEP-0136 that is intentionally not defined by this specification for simplicity:
An archive contains a collection of messages relevant to a particular XMPP address, e.g. a user, MUC, pubsub node, server. Note: while a service might have many "archives" as defined here (one per JID capable of being queried) this is a conceptual distinction, and a server is not bound to any particular implementation or arrangement of data stores.
Exactly which messages a server archives is up to implementation and deployment policy, but it is expected that all messages that hold meaningful content, rather than state changes such as Chat State Notifications, would be archived. Rules are specified later in this document.
A stored message consists of at least the following pieces of information:
Note that 'incoming' and 'outgoing' messages are viewed within the context of the archived JID, rather than the system as a whole. For example, if email@example.com sent a message to firstname.lastname@example.org, it would be an outgoing message in the context of archiving for Romeo, and an incoming message in the context of archiving for Juliet.
Order within the archive MUST be preserved, where the order of messages is the same as the order that the client originally received them (or would have received them if online). Throughout this document the term 'chronological order' refers to this order, however implementors should take care not to rely on timestamps alone for ordering messages, as multiple messages may share the same timestamp.
A server MAY impose limits on the size of an individual archive. For example a server might begin to discard old messages once the archive reaches a certain size, or only keep messages until they reach a certain age. Any such deleted messages MUST be the oldest in the archive, i.e. it is not permitted to create gaps or "holes" in the archive. The UIDs of deleted messages MUST NOT be reused for new messages.
However a server that wishes to remove messages from the middle of an archive, e.g. to remove accidentally transmitted sensitive information may omit the <message> stanza from inside the <forwarded> element or replace the message with an appropriate placeholder when transmitting the result in response to a query. However servers MUST retain the UID, timestamp and JID of the original message internally to ensure that all queries remain consistent. It should also be understood that clients maintaining their own local copy of the archive may still retain the original message locally in this case, and this protocol provides no mechanism for forcibly removing messages from any local archive or cache that clients may keep.
There is no restriction on which services can expose archives, although only user and MUC archives are discussed here.
The most typical address is that of a user's own bare JID, within which those messages sent to or from that user's account would generally automatically be stored by the server. The collection is ordered chronologically by the time each message was sent/received.
Servers that expose archive messages of sent/received messages on behalf of local users MUST expose these archives to the user on the user's bare JID.
A MUC service allowing MAM queries for a room MUST expose the MAM archive on the room's bare JID
While this document talks about 'clients' and 'servers', as these are the common cases, the querying entity (referred to as a 'client') need not be an XMPP client as defined by RFC6120, but could potentially be any type of entity, and the queried entity (referred to as a 'server') need not be an XMPP server as defined by RFC6120, although access controls might prohibit any given entity from being able to access an archive.
When a message is archived, the server MUST add an <stanza-id/> element as defined in Unique and Stable Stanza IDs (XEP-0359)  to the message, which informs the recipient of where and under what ID the message is stored. When doing this the server MUST follow the business rules defined in XEP-0359. The 'by' attribute MUST be set to the address of the archive. For regular users that’s the bare JID of the account and for MUC that’s the bare JID of the room.
Servers MUST NOT include the <stanza-id/> element in messages addressed to JIDs that do not have permissions to access the archive, such as a users’s outgoing messages to their contacts. However servers SHOULD include the element as a child of the forwarded message when using Message Carbons (XEP-0280) 
Note: Previous versions of this protocol did not specify any interaction with stanza-id, and clients MUST NOT interpret XEP-0359 IDs in messages as archive IDs unless the server advertises support for 'urn:xmpp:mam:2' specifically.
An entity is able to query (subject to appropriate access rights) an archive for all messages within a certain timespan, optionally restricting results to those to/from a particular JID. To allow limiting the results or paging through them a client may use Result Set Management (XEP-0059) , which MUST be supported by both the client and the server.
A query consists of an <iq/> stanza of type 'set' addressed to the account or server entity hosting the archive, with a 'query' payload. On receiving the query, the server pushes to the client a series of messages in chronological order from the archive that match the client's given criteria. After the results it then returns the <iq/> result to indicate that the query is completed.
The final <iq/> result response MUST include an RSM <set/> element, wrapped into a <fin/> element qualified by the 'urn:xmpp:mam:2' namespace, indicating the UID of the first and last message of the (possibly limited) result set. This allows clients to accurately page through messages.
To ensure that the client knows when the results are complete, the server MUST send the <iq/> result after last query result has been sent to the client. The client can optionally include a 'queryid' attribute in their query, which allows the client to match results to their initiating query.
By default all messages match a query, and filters are used to request a subset of the archived messages. Filters are specified in a Data Forms (XEP-0004)  data form included with the query. The hidden FORM_TYPE field MUST be set to this protocol's namespace, 'urn:xmpp:mam:2'. Six further fields are defined by this XEP and MUST be supported by servers, though all of them are optional for the client. These fields are:
Servers supporting fields marked with an asterisk (*) MUST advertise the disco feature 'urn:xmpp:mam:2#extended' and clients that depend on these fields MUST verify that the server advertises this feature before attempting to use them.
Other fields may be used, but are not defined in this document - the naming of new fields MUST be consistent with the format defined in Field Standardization for Data Forms (XEP-0068) . Servers MUST NOT mark any fields in the form as being required (i.e. with the data forms <required/> element), regardless of whether they are defined in this document or elsewhere.
If a 'with' field is present in the form, it contains a JID against which to match messages. The server MUST only return messages if they match the supplied JID. A message in a user's archive matches if the JID matches either the to or from of the message. An item in a MUC archive matches if the publisher of the item matches the JID; note that this should only be available to entities that would already have been allowed to know the publisher of the events (e.g. this could not be used by a visitor to a semi-anonymous MUC).
To allow querying for messages the user sent to themselves, the client needs to set the 'with' attribute to the account JID. In that case, the server MUST only return results where both the 'to' and 'from' match the bare JID (either as bare or by ignoring the resource), as otherwise every message in the archive would match.
If 'with' is omitted, the server MUST match all messages in the selected timespan with the query, regardless of the to/from addresses on each message.
If (and only if) the supplied JID is a bare JID (i.e. no resource is present), then the server SHOULD return messages if their bare to/from address for a user archive, or from address otherwise, would match it. For example, if the client supplies a 'with' of "email@example.com" a query to their own archive would also match messages to or from "firstname.lastname@example.org/balcony" and "email@example.com/chamber".
The 'start' field is used to filter out messages before a certain date/time. If specified, a server MUST only return messages whose timestamp is equal to or later than the given timestamp.
If omitted, the server SHOULD assume the value of 'start' to be equal to the date/time of the earliest message stored in the archive.
Conversely, the 'end' field is used to exclude from the results messages after a certain point in time. If specified, a server MUST only return messages whose timestamp is equal to or earlier than the timestamp given in the 'end' field.
If omitted, the server SHOULD assume the value of 'end' to be equal to the date/time of the most recent message stored in the archive.
If the client has already seen some messages, it may choose to restrict its query to before and/or after messages it already knows about. This may be done through the 'before-id' and 'after-id' fields.
If the client already knows the UID of one or more messages it wants to fetch, it can use the 'ids' field:
If any UID requested by the client in any of the 'before-id', 'after-id' or 'ids' form fields is not present in the archive, the server MUST return an item-not-found error in response to the query.
In order for the client find out about additional fields the server might support, it can send an iq stanza of type 'get' addressed to the archive like this:
The server replies with all the form fields it supports in queries, which MUST include the mandatory fields specified in this document.
If the client understands any of the additional fields it MAY proceed to include any of them in subsequent queries. It is not required to include any or all of the supported fields in queries.
A special note about the 'ids' field: this field is of type 'list-multi' which typically is used to allow the client to select from a provided list of options. In this case the list of all possible ids MUST NOT be provided by the server, as it is likely to be extremely large. Instead the server MUST include a Data Forms Validation (XEP-0122)  <validate/> element that signals the list is open to arbitrary values provided by the client.
Note that as the 'with', 'start' and 'end' fields MUST be implemented by servers, clients are able to submit forms using combinations of only these fields without needing to first fetch the form from the server and the types of these fields MUST be 'jid-single', 'text-single' and 'text-single' respectively. A server MUST NOT rely on a client having first requested the form before submitting queries
If a client includes a form field that the server does not recognise, the server MUST respond with a 'feature-not-implemented' error.
The server responds to the archive query by transmitting to the client all the messages that match the criteria the client requested, subject to implementation limits. The results are sent as individual stanzas, with the original message encapsulated in a <forwarded/> element as described in Stanza Forwarding (XEP-0297) .
The result messages MUST contain a <result/> element with an 'id' attribute that gives the current message's archive UID (archived messages MAY also contain a XEP-0359 <stanza-id> element, but clients MUST NOT depend on it). If the client gave a 'queryid' attribute in its initial query, the server MUST also include that in this result element.
The <result/> element contains a <forwarded/> element which SHOULD contain the original message as it was received, and SHOULD also contain a <delay/> element qualified by the 'urn:xmpp:delay' namespace specified in Delayed Delivery (XEP-0203) . The value of the 'stamp' attribute MUST be the time the message was originally received by the forwarding entity.
The archive results MUST be sorted in chronological order, both within the returned results and within the ordering of RSM such that if a client were to request the first 10 stanzas in an archive, then use RSM to request the next 10 stanzas (by providing the 'after' element with the UID of the 10th stanza in the first results) all 20 result stanzas would be received in chronological order.
A client or server will typically want to limit the number of results transmitted at a time, thereby breaking the result stream into smaller 'pages'. For this purpose a server MUST support Result Set Management (XEP-0059)  and MUST support the paging mechanism defined therein. A client MAY include a <set/> element in its query.
For the purposes of this protocol, the UIDs used by RSM correspond with the UIDs of the stanzas stored in the archive.
To conserve resources, a server MAY place a reasonable limit on how many stanzas may be pushed to a client in one request. Whether or not the client query included a <set/> element, the server MAY simply return its limited results, modifying the <set/> element it returns appropriately.
The <first> and <last> elements specify the UID of the first and last returned results (not necessarily of all the messages that matched the query, if the results have been limited).
The RSM <count> element and the 'index' attribute on the RSM <first> element are optional, a server MAY include them, but a client MUST NOT depend on them being present. Please refer to the RSM specification for more information surrounding their meaning and use.
Having previously made a query that returned results limited by the server (as described above), a client can re-send the same request and receive the next 'page' of results. It does this by including a <set> element with its request, containing an <after/> with the UID of the last message it received from the previous query.
Note: There is no concept of an "open query", and servers MUST be prepared to receive arbitrary page requests at any time.
RSM does not define the behaviour of including both <before> and <after> in the same request. To retrieve a range of items between two known ids, use before-id and after-id in the query form instead.
If the UID contained within an <after> or <before> element is not present in the archive, the server MUST return an item-not-found error in response to the query.
When the results returned by the server are complete (that is: when they have not been limited by the maximum size of the result page (either as specified or enforced by the server)), the server MUST include a 'complete' attribute on the <fin> element, with a value of 'true'; this informs the client that it doesn't need to perform further paging to retreive the requested data. If it is not the last page of the result set, the server MUST either omit the 'complete' attribute, or give it a value of 'false'.
Sometimes (e.g. due to network or storage partitioning, or other transient errors) the server might return results to a client that are unstable (e.g. they might later change in sequence or content). In such a situation the server MUST stamp the <fin> element with a 'stable' attribute with a value of 'false'. If the server knows that the data it's serving are stable it MUST either stamp a 'stable' attribute with a value of 'true', or no such attribute. An example of when unstable might legitimately be returned is if the MAM service uses a clustered data store and a query covers a time period for which the data store has not yet converged; it the server could return best-guess results and tell the client that they may be unstable. A client SHOULD NOT cache unstable results long-term without later confirming (by reissuing appropriate queries) that they have become stable.
To request the page at the end of the archive (i.e. the most recent messages), include just an empty <before/> element in the RSM part of the query. As defined by RSM, this will return the last page of the archive.
Within the returned page, all results are still in chronological order, that is, the first result you receive will be the oldest item in the page, and the last result you receive will be the last item in the archive.
When fetching a page, the client may prefer for the server to send the results within that page in reverse order. For example, if a client implements a user interface that automatically fetches older messages as a user scrolls backward, it may want to receive and display the newest messages first, instead of waiting for the whole page to be received.
A client wishing for a reversed page should include the <flip-page/> element in its query, like so:
It is important to note that flipping a page does not affect what results are returned in response to the query. It only affects the order in which they are transmitted from the server to the client.
A client that wishes to use flipped pages MUST ensure that the server advertises the 'urn:xmpp:mam:2#extended' feature.
When planning a query, a client may wish to learn the current state of the archive. This includes information about the first/last entries in the archive.
When the archive advertises support for 'urn:xmpp:mam:2#extended' then the archive supports queries for this metadata via an iq of type 'get' to the archive's address, with a <metadata/> payload in the 'urn:xmpp:mam:2' namespace.
The server response includes a <metadata/> element containing information about the archive. This element MUST include <start/> and <end/> elements, which each have an 'id' and XEP-0082 formatted 'timestamp of the first and last messages in the archive respectively.
Different entities will have different requirements for which data are stored, as might different deployments. This section provides general rules within which a server will act. While there may be local policy restrictions that prevent archiving of some aspects discussed here, this is a RECOMMENDED baseline. A server MAY implement any subset of possible archives for JIDs it controls (although it MUST advertise support only for those JIDs that support it).
No requirements are placed on how a server implements its storage beyond that it has to store data sufficient to be able to comply with this document. When this document describes storage requirements (e.g. MUST NOT store more than one copy...), it refers to what would appear to have been stored in order to satisfy the query.
If an entity (user's server, MUC room, pubsub node, ...) rejects an incoming message (such as from an occupant not allowed to send messages to the room, a user not authorized to publish to a pubsub node, a contact blocked by the user etc.) that message should not appear in the archive for the entity that rejected it - the archive should represent what logical entities (MUC occupants, users, pubsub subscribers...) would have received, and so only contain messages accepted for delivery to such entities.
A user archive is anticipated to provide the user with the ability to access their prior conversations. To this end, a server SHOULD include in a user archive all of the messages a user sends or receives of type 'normal' or 'chat' that contain a <body> element. A server SHOULD also include messages of type 'groupchat' that have a <body>, but where such history is accessible through another method (e.g. through an archive on the MUC JID), a server MAY exclude these from the archive. A server MAY include additional non-conversation messages. A server MAY include messages of type 'headline', but this is not generally suggested.
At a minimum, the server MUST store the <body> elements of a stanza. It is suggested that other elements that are used in a given deployment to supplement conversations (e.g. XHTML-IM payloads) are also stored. Other elements MAY be stored.
If a server supports mechanisms that multiply copies of a stanza (e.g. Carbons, or forking a stanza to a bare JID), it MUST store such a staza within a given archive only once, irrespective of multiple connected clients receiving copies
A MUC archives allows a user to view the conversation within a room. All messages sent to the room that contain a <body> element SHOULD be stored, as should subject change stanzas, apart from those messages that the room rejects.
A MUC archive MUST store each message only once (not, for example, every copy sent out to an occupant).
A MUC archive MUST NOT include 'private message' results (those sent directly between occupants, not shared in the room) in the results
A MUC archive MUST check that the user requesting the archive has the right to enter it at the time of the query and only allow access if so. In a members-only chat room, only owners, admins or members can query a room archive. In the case of open MUC rooms, the MUC archives can generally be accessed by any users (including those who have never entered the room) who do not have an affiliation of 'outcast', but a MUC archive MAY further limit access based on other criteria as part of the deployment policy. A MUC archive MAY, if it stores historical data about previous configuration states, limit the results returned to only those that the querying user would have been authorised to see at the time (e.g. it MAY limit the results to not include results while a user was an outcast).
When sending out the archives to a requesting client, the forwarded stanza MUST NOT have a 'to' attribute, and the 'from' MUST be the occupant JID of the sender of the archived message.
In the case of non-anonymous rooms or if the recipient of the MUC archive has the right to access the sender real JID at the time of the query, the archive message will use extended message information in an <x/> element qualified by the 'http://jabber.org/protocol/muc#user' namespace and containing an <item/> child with a 'jid' attribute specifying the occupant's full JID, as defined for non-anonymous room presence in Multi-User Chat (XEP-0045) . The archiving entity MUST strip any pre-existing <x> element from MUC messages (as MUC rooms are not required to do this).
This specification reserves the 'node' attribute of the <query> element for use with pubsub archives. Full details of using this protocol to query pubsub node archives are documented in [FIXME: number to be assigned to xep-pubsub-mam].
The IDs used within an archive MUST be unique per item stored and MUST NOT be reused, even if the original item with a given ID has since been removed from the archive. If a server provides multiple archives (e.g. many user archives, or many MUC archives), the IDs do not need to be unique across all of these archives unless the server also allows a single query to be run across multiple archives (e.g. searching of all MUC rooms), discussion of which is beyond the scope of this document. These IDs are strings that servers may construct in any manner, and clients must treat as opaque strings (e.g. there is no requirement for them to be numeric, sequenced or GUIDs).
In addition to one-off queries, clients may use this protocol to synchronize a local archive with the server's archive. However because this protocol is detached from normal routing of messages, it is possible that a client will receive messages while trying to synchronize, which has the potential to cause duplicated messages. Resolving this is beyond the scope of this specification, but may instead be solved during the initial connection phase by using an alternative connection protocol such as Bind 2.0 (XEP-0386) .
If a server or other entity hosts archives and supports MAM queries, it MUST advertise the 'urn:xmpp:mam:2' and 'urn:xmpp:mam:2#extended' features in response to Service Discovery (XEP-0030)  requests made to archiving JIDs (i.e. JIDs hosting an archive, such as users' bare JIDs):
Servers advertising the 'urn:xmpp:mam:2#extended' feature MUST implement the 'before-id' and 'after-id' fields, as well as support for flipped pages and single-item retrieval. The 'urn:xmpp:mam:2#extended' feature MUST NOT be advertised by a server without also advertising 'urn:xmpp:mam:2'.
|Queries using 'with', 'start' and 'end'||Required||Required|
|Error responses for missing UIDs||Required||Required|
|Queries using 'before-id', 'after-id' or 'ids'||-||Required|
|Archive metadata query||-||Required|
An archive generally consists of private conversations, and so a server MUST adequately protect an archive from unauthorized third-party access. For example authorized parties for a user's archive would likely include just the user, and a MUC archive for a private room might be restricted to room members. An implementation MAY choose to allow access to any archive by server administrators. If a client requests access to an archive it does not have permissions for the server MUST return an iq with type error, and the error condition SHOULD be 'forbidden'.
A server SHOULD provide a mechanism for a user to disable archiving of messages with all or specific contacts, such as via the configuration protocol described in this document. This allows the user to prevent the archiving of potentially sensitive messages in the first place.
A server MAY automatically prevent certain sensitive messages from being archived. How such messages are identified is beyond the scope of this specification, but technologies such as Security Labels in XMPP (XEP-0258)  may be used, for example.
Entities that implement this specification must also adhere to the security requirements of XEP-0359.
This specification re-uses the <x> element from the 'http://jabber.org/protocol/muc#user' namespace to convey information about the sender of a message in a MUC room. However this element is not sanitized by MUC services, so the archiving entity MUST strip any existing <x> element in the 'http://jabber.org/protocol/muc#user' namespace from messages before archiving them (regardless of whether it adds in its own <x> element).
Many thanks to Dave Cridland, Kim Alvefur, Yann Leboulanger, Evgeny Khramtsov, Florian Schmaus, Lance Stout, Waqas Hussain, Daniel Gultsch, Philipp Hörist, Jonas Schäfer and Georg Lukas for their input and feedback on this specification.
This document in other formats: XML PDF
This XMPP Extension Protocol is copyright © 1999 – 2020 by the XMPP Standards Foundation (XSF).
Permission is hereby granted, free of charge, to any person obtaining a copy of this specification (the "Specification"), to make use of the Specification without restriction, including without limitation the rights to implement the Specification in a software program, deploy the Specification in a network service, and copy, modify, merge, publish, translate, distribute, sublicense, or sell copies of the Specification, and to permit persons to whom the Specification is furnished to do so, subject to the condition that the foregoing copyright notice and this permission notice shall be included in all copies or substantial portions of the Specification. Unless separate permission is granted, modified works that are redistributed shall not contain misleading information regarding the authors, title, number, or publisher of the Specification, and shall not claim endorsement of the modified works by the authors, any organization or project to which the authors belong, or the XMPP Standards Foundation.
## NOTE WELL: This Specification is provided on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, express or implied, including, without limitation, any warranties or conditions of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A PARTICULAR PURPOSE. ##
In no event and under no legal theory, whether in tort (including negligence), contract, or otherwise, unless required by applicable law (such as deliberate and grossly negligent acts) or agreed to in writing, shall the XMPP Standards Foundation or any author of this Specification be liable for damages, including any direct, indirect, special, incidental, or consequential damages of any character arising from, out of, or in connection with the Specification or the implementation, deployment, or other use of the Specification (including but not limited to damages for loss of goodwill, work stoppage, computer failure or malfunction, or any and all other commercial damages or losses), even if the XMPP Standards Foundation or such author has been advised of the possibility of such damages.
This XMPP Extension Protocol has been contributed in full conformance with the XSF's Intellectual Property Rights Policy (a copy of which can be found at <https://xmpp.org/about/xsf/ipr-policy> or obtained by writing to XMPP Standards Foundation, P.O. Box 787, Parker, CO 80134 USA).
The Extensible Messaging and Presence Protocol (XMPP) is defined in the XMPP Core (RFC 6120) and XMPP IM (RFC 6121) specifications contributed by the XMPP Standards Foundation to the Internet Standards Process, which is managed by the Internet Engineering Task Force in accordance with RFC 2026. Any protocol defined in this document has been developed outside the Internet Standards Process and is to be understood as an extension to XMPP rather than as an evolution, development, or modification of XMPP itself.
The primary venue for discussion of XMPP Extension Protocols is the <firstname.lastname@example.org> discussion list.
Discussion on other xmpp.org discussion lists might also be appropriate; see <http://xmpp.org/about/discuss.shtml> for a complete list.
Errata can be sent to <email@example.com>.
The following requirements keywords as used in this document are to be interpreted as described in RFC 2119: "MUST", "SHALL", "REQUIRED"; "MUST NOT", "SHALL NOT"; "SHOULD", "RECOMMENDED"; "SHOULD NOT", "NOT RECOMMENDED"; "MAY", "OPTIONAL".
Note: Older versions of this specification might be available at http://xmpp.org/extensions/attic/
Fix missing part of sentence to make more sense
Add 'before-id' and 'after-id' fields, flipped pages, single-item retrieval and a new mandatory disco feature
Split preferences protocol into a separate document
Split the details of pubsub archives into a separate document
Warn of spoofing of <x> elements in MUC stanzas
Namespace bump from mam:1 to mam:2
Added method for server to communicate the archive id
Incorporated editorial clarifications based on implementation feedback
Clarifications on the topics of message ordering, message deletion and use of the protocol for synchronization
Bump old namespace version in examples (Kevin Smith)
Added clarifications and enhancements when interacting with Multi-User Chat (MUC) (Mickaël Rémond)
Fix Example #9 from </message/> to </iq> (Florian Schmaus)
Add missing jabber:client namespace in Example #2 (Christian Schudt)
Switch the sentinel message back to the iq result.
Various small fixes to the document.
Fetching current preferences, switch to iq-set for searching, switch to using a data form, describe how to fetch that form, remove the archived element and use a sentinel message instead of iq reply.
Document the ability to page through results by message UIDs, define the <archived/> element, and various minor improvements.
Initial version, to much rejoicing.