&xep0096; was the original XMPP protocol extension for file transfer negotiation. However, that protocol has several drawbacks, most related to the &xep0095; protocol on which it depends:

1. It does not enable a true, bidirectional negotiation; instead, the initiator sets the terms for the file transfer and the responder either accepts the terms or cancels the negotiation.
2. It is the only technology in the Jabber/XMPP protocol "stack" that uses XEP-095: Stream Initiation. More modern technologies such as voice and video session negotiation use &xep0166;, and it would be helpful if implementors could re-use the same code for all negotiation use cases.

To overcome these drawbacks, this specification defines a file transfer negotiation method that meets the following requirements:

• Reuse the session negotiation semantics from XEP-0166.
• Reuse the file description format from XEP-0096.
• Define a clear upgrade path from XEP-0096 to this specification.

Jingle file transfer is only as reliable as the transports on which it depends. In particular, SOCKS5 Bytestreams ("S5B") does not always result in NAT or firewall traversal. To work around that problem, this specification requires all implementations to support as a fallback mechanism In-Band Bytestreams ("IBB"), which usually results in a successful (if slow) file transfer.

Note: It is likely that a future version of this specification will also recommend implementation of a Jingle transport method that emulates the IETF's ICE-TCP technology, which is currently a work in progress (see &ice-tcp;); however, a future Jingle ICE-TCP transport method is dependent on the outcome of IETF work in this area.

This section provides a friendly introduction to Jingle file transfer.

First, the party that wishes to initiate the file transfer determines the responder's capabilities (via &xep0030; or &xep0115;). Here we assume that the responder supports the following service discovery features:

• urn:xmpp:jingle:1 as described in XEP-0166
• urn:xmpp:jingle:apps:file-transfer:1 as defined in this document &NSVER;
• urn:xmpp:jingle:transports:s5b:1 as defined in &xep0260;
• urn:xmpp:jingle:transports:ibb:0 as defined in &xep0261;

The initiator then sends a Jingle session-initiation request to a potential responder. The content-type of the request specifies two things:

1. An application type of "urn:xmpp:jingle:apps:file-transfer:1". In particular, the <description/> element contains an <offer/> or <request/> element that in turn contains a <file/> element qualified by the existing 'http://jabber.org/protocol/si/profile/file-transfer' namespace from XEP-0096.
2. An appropriate transport method. So far the suggested methods are jingle-s5b (XEP-0260) and, as a fallback, jingle-ibb (XEP-0261).

Note: All attributes of the <file/> element are defined in XEP-0096, not in this specification.

In this example, the initiator is <romeo@montague.lit>, the responder is <juliet@capulet.lit>, the application type is a file offer, and the transport method is jingle-s5b.

The flow is as follows.

First the initiator sends a Jingle session-initiate.

Note: Inclusion of the <range/> child of the <file/> element indicates that the initiatior supports ranged transfers as described below under Ranged Transfers.

Note: Computing the hash of the file before sending it can slow down the process of file transfer, since the sending application needs to process the file twice. The sender might prefer to send the hash after the file transfer has begun, using a transport-info message as described under Communicating the Hash.

The responder immediately acknowledges receipt of the Jingle session-initiate.

The initiator then attempts to initiate a SOCKS5 Bytestream with the responder as described in XEP-0260 and XEP-0065.

If the responder is able to connect to one of the streamhosts, it returns a Jingle session-accept (including only the JID of the streamhost to which it connected).

Note: Inclusion of the <range/> child of the <file/> element indicates that the receiver also supports ranged transfers as described below under Ranged Transfers.

The initiator acknowledges the Jingle session-accept.

Now the parties exchange the file using the negotiated transport (here, SOCKS5 Bytestreams).

Once the transfer is completed, either party can terminate the Jingle session; preferably this is done by the entity that receives the file to ensure that the complete file (up to the advertised size) has been received.

For a description of the transport fallback scenario (from SOCK5 Bytestreams to In-Band Bytestreams), refer to XEP-0260.

At any time during the lifetime of the file transfer session, the hosting entity can communicate the hash of the file to the receiving entity. This is done by sending a session-info message containing a <hash/> element qualified by the 'urn:xmpp:jingle:apps:file-transfer:info:1' namespace. The XML character data of this <hash/> element has the same meaning as the 'hash' attribute of the <file/> element qualified by the 'http://jabber.org/protocol/si/profile/file-transfer' namespace from XEP-0096; that is, it is a checksum of the file contents produced in accordance with the hashing function specified by the 'algo' attribute, which MUST be one of the functions listed in the &ianahashes;. For historical reasons and for backward-compatibility with XEP-0096, the 'algo' attribute defaults to a value of "md5".

If the entity that hosts a file has advertised its existence (or if a previous file transfer attempt has failed and the receiver would like to initiate another attempt), the entity that wishes to receive the file can "pull" the file from the hosting entity. This is done by sending a Jingle session-initiate to the hosting entity, including a &DESCRIPTION; element qualified by the 'urn:xmpp:jingle:apps:file-transfer:1' namespace and containing a <request/> element that defines the requested file.

The parties would then complete a session negotiation flow similar to that outlined above for offering a file.

Note: If the requesting entity knows the hash of the file, then it can include only that metadata in its request. If not, the requesting entity needs to include enough metadata to uniquely identify the file, such as the date, name, and size. For similar considerations, see &rfc5547;.

As in XEP-0096, a transfer can include only part of a file (e.g., to restart delivery of a truncated transfer session at a point other than the start of the file). This is done using the <range/> element from XEP-0096. The usage is illustrated in the following examples.

Let us imagine that the parties negotiate a file transfer session using, say, In-Band Bytestreams. During the transfer, the recipient goes offline unexpectedly and IBB stanzas from the sender to the recipient begin to bounce. When the recipient comes back online, the recipient could initiate a new Jingle session (to retrieve the file) and specify that it wants to receive all chunks after byte 270336 (which might be the 66th chunk of size 4096).

Alternatively, the sender could initiate a new file transfer, indicating that it supports ranged transfers, and in the Jingle session-accept message the receiver could indicate that it wants the transfer to begin at the specified offset.

Jingle file transfer uses only a few of the actions defined in XEP-0166. Jingle usage is summarized in the following table.

For historical reasons and for backward-compatibility with XEP-0096, the hashing algorithm used in computing the file checksum defaults to MD5. It is RECOMMENDED for implementations to use stronger hashing algorithms.

In order to secure the data stream, implementations SHOULD use encryption methods appropriate to the transport method being used. For example, end-to-end encryption can be negotiated over either SOCKS5 Bytestreams or In-Band Bytestreams as described in XEP-0260 and XEP-0261.

Refer to XEP-0047, XEP-0065, XEP-0096, XEP-0260, and XEP-0261 for related security considerations.

No interaction with &IANA; is required as a result of this document.

Thanks to Steffen Larsen, Marcus Lundblad, Joe Maissel, Ali Sabil, Matthew Wild, and Jiří Zárevúcky for their feedback.