Documentation / technical / protocol-v2.txton commit general improvements (43abf13)
   1Git Wire Protocol, Version 2
   2============================
   3
   4This document presents a specification for a version 2 of Git's wire
   5protocol.  Protocol v2 will improve upon v1 in the following ways:
   6
   7  * Instead of multiple service names, multiple commands will be
   8    supported by a single service
   9  * Easily extendable as capabilities are moved into their own section
  10    of the protocol, no longer being hidden behind a NUL byte and
  11    limited by the size of a pkt-line
  12  * Separate out other information hidden behind NUL bytes (e.g. agent
  13    string as a capability and symrefs can be requested using 'ls-refs')
  14  * Reference advertisement will be omitted unless explicitly requested
  15  * ls-refs command to explicitly request some refs
  16  * Designed with http and stateless-rpc in mind.  With clear flush
  17    semantics the http remote helper can simply act as a proxy
  18
  19In protocol v2 communication is command oriented.  When first contacting a
  20server a list of capabilities will advertised.  Some of these capabilities
  21will be commands which a client can request be executed.  Once a command
  22has completed, a client can reuse the connection and request that other
  23commands be executed.
  24
  25Packet-Line Framing
  26-------------------
  27
  28All communication is done using packet-line framing, just as in v1.  See
  29`Documentation/technical/pack-protocol.txt` and
  30`Documentation/technical/protocol-common.txt` for more information.
  31
  32In protocol v2 these special packets will have the following semantics:
  33
  34  * '0000' Flush Packet (flush-pkt) - indicates the end of a message
  35  * '0001' Delimiter Packet (delim-pkt) - separates sections of a message
  36
  37Initial Client Request
  38----------------------
  39
  40In general a client can request to speak protocol v2 by sending
  41`version=2` through the respective side-channel for the transport being
  42used which inevitably sets `GIT_PROTOCOL`.  More information can be
  43found in `pack-protocol.txt` and `http-protocol.txt`.  In all cases the
  44response from the server is the capability advertisement.
  45
  46Git Transport
  47~~~~~~~~~~~~~
  48
  49When using the git:// transport, you can request to use protocol v2 by
  50sending "version=2" as an extra parameter:
  51
  52   003egit-upload-pack /project.git\0host=myserver.com\0\0version=2\0
  53
  54SSH and File Transport
  55~~~~~~~~~~~~~~~~~~~~~~
  56
  57When using either the ssh:// or file:// transport, the GIT_PROTOCOL
  58environment variable must be set explicitly to include "version=2".
  59
  60HTTP Transport
  61~~~~~~~~~~~~~~
  62
  63When using the http:// or https:// transport a client makes a "smart"
  64info/refs request as described in `http-protocol.txt` and requests that
  65v2 be used by supplying "version=2" in the `Git-Protocol` header.
  66
  67   C: GET $GIT_URL/info/refs?service=git-upload-pack HTTP/1.0
  68   C: Git-Protocol: version=2
  69
  70A v2 server would reply:
  71
  72   S: 200 OK
  73   S: <Some headers>
  74   S: ...
  75   S:
  76   S: 000eversion 2\n
  77   S: <capability-advertisement>
  78
  79Subsequent requests are then made directly to the service
  80`$GIT_URL/git-upload-pack`. (This works the same for git-receive-pack).
  81
  82Capability Advertisement
  83------------------------
  84
  85A server which decides to communicate (based on a request from a client)
  86using protocol version 2, notifies the client by sending a version string
  87in its initial response followed by an advertisement of its capabilities.
  88Each capability is a key with an optional value.  Clients must ignore all
  89unknown keys.  Semantics of unknown values are left to the definition of
  90each key.  Some capabilities will describe commands which can be requested
  91to be executed by the client.
  92
  93    capability-advertisement = protocol-version
  94                               capability-list
  95                               flush-pkt
  96
  97    protocol-version = PKT-LINE("version 2" LF)
  98    capability-list = *capability
  99    capability = PKT-LINE(key[=value] LF)
 100
 101    key = 1*(ALPHA | DIGIT | "-_")
 102    value = 1*(ALPHA | DIGIT | " -_.,?\/{}[]()<>!@#$%^&*+=:;")
 103
 104Command Request
 105---------------
 106
 107After receiving the capability advertisement, a client can then issue a
 108request to select the command it wants with any particular capabilities
 109or arguments.  There is then an optional section where the client can
 110provide any command specific parameters or queries.  Only a single
 111command can be requested at a time.
 112
 113    request = empty-request | command-request
 114    empty-request = flush-pkt
 115    command-request = command
 116                      capability-list
 117                      [command-args]
 118                      flush-pkt
 119    command = PKT-LINE("command=" key LF)
 120    command-args = delim-pkt
 121                   *command-specific-arg
 122
 123    command-specific-args are packet line framed arguments defined by
 124    each individual command.
 125
 126The server will then check to ensure that the client's request is
 127comprised of a valid command as well as valid capabilities which were
 128advertised.  If the request is valid the server will then execute the
 129command.  A server MUST wait till it has received the client's entire
 130request before issuing a response.  The format of the response is
 131determined by the command being executed, but in all cases a flush-pkt
 132indicates the end of the response.
 133
 134When a command has finished, and the client has received the entire
 135response from the server, a client can either request that another
 136command be executed or can terminate the connection.  A client may
 137optionally send an empty request consisting of just a flush-pkt to
 138indicate that no more requests will be made.
 139
 140Capabilities
 141------------
 142
 143There are two different types of capabilities: normal capabilities,
 144which can be used to convey information or alter the behavior of a
 145request, and commands, which are the core actions that a client wants to
 146perform (fetch, push, etc).
 147
 148Protocol version 2 is stateless by default.  This means that all commands
 149must only last a single round and be stateless from the perspective of the
 150server side, unless the client has requested a capability indicating that
 151state should be maintained by the server.  Clients MUST NOT require state
 152management on the server side in order to function correctly.  This
 153permits simple round-robin load-balancing on the server side, without
 154needing to worry about state management.
 155
 156agent
 157~~~~~
 158
 159The server can advertise the `agent` capability with a value `X` (in the
 160form `agent=X`) to notify the client that the server is running version
 161`X`.  The client may optionally send its own agent string by including
 162the `agent` capability with a value `Y` (in the form `agent=Y`) in its
 163request to the server (but it MUST NOT do so if the server did not
 164advertise the agent capability). The `X` and `Y` strings may contain any
 165printable ASCII characters except space (i.e., the byte range 32 < x <
 166127), and are typically of the form "package/version" (e.g.,
 167"git/1.8.3.1"). The agent strings are purely informative for statistics
 168and debugging purposes, and MUST NOT be used to programmatically assume
 169the presence or absence of particular features.
 170
 171ls-refs
 172~~~~~~~
 173
 174`ls-refs` is the command used to request a reference advertisement in v2.
 175Unlike the current reference advertisement, ls-refs takes in arguments
 176which can be used to limit the refs sent from the server.
 177
 178Additional features not supported in the base command will be advertised
 179as the value of the command in the capability advertisement in the form
 180of a space separated list of features: "<command>=<feature 1> <feature 2>"
 181
 182ls-refs takes in the following arguments:
 183
 184    symrefs
 185        In addition to the object pointed by it, show the underlying ref
 186        pointed by it when showing a symbolic ref.
 187    peel
 188        Show peeled tags.
 189    ref-prefix <prefix>
 190        When specified, only references having a prefix matching one of
 191        the provided prefixes are displayed.
 192
 193The output of ls-refs is as follows:
 194
 195    output = *ref
 196             flush-pkt
 197    ref = PKT-LINE(obj-id SP refname *(SP ref-attribute) LF)
 198    ref-attribute = (symref | peeled)
 199    symref = "symref-target:" symref-target
 200    peeled = "peeled:" obj-id
 201
 202fetch
 203~~~~~
 204
 205`fetch` is the command used to fetch a packfile in v2.  It can be looked
 206at as a modified version of the v1 fetch where the ref-advertisement is
 207stripped out (since the `ls-refs` command fills that role) and the
 208message format is tweaked to eliminate redundancies and permit easy
 209addition of future extensions.
 210
 211Additional features not supported in the base command will be advertised
 212as the value of the command in the capability advertisement in the form
 213of a space separated list of features: "<command>=<feature 1> <feature 2>"
 214
 215A `fetch` request can take the following arguments:
 216
 217    want <oid>
 218        Indicates to the server an object which the client wants to
 219        retrieve.  Wants can be anything and are not limited to
 220        advertised objects.
 221
 222    have <oid>
 223        Indicates to the server an object which the client has locally.
 224        This allows the server to make a packfile which only contains
 225        the objects that the client needs. Multiple 'have' lines can be
 226        supplied.
 227
 228    done
 229        Indicates to the server that negotiation should terminate (or
 230        not even begin if performing a clone) and that the server should
 231        use the information supplied in the request to construct the
 232        packfile.
 233
 234    thin-pack
 235        Request that a thin pack be sent, which is a pack with deltas
 236        which reference base objects not contained within the pack (but
 237        are known to exist at the receiving end). This can reduce the
 238        network traffic significantly, but it requires the receiving end
 239        to know how to "thicken" these packs by adding the missing bases
 240        to the pack.
 241
 242    no-progress
 243        Request that progress information that would normally be sent on
 244        side-band channel 2, during the packfile transfer, should not be
 245        sent.  However, the side-band channel 3 is still used for error
 246        responses.
 247
 248    include-tag
 249        Request that annotated tags should be sent if the objects they
 250        point to are being sent.
 251
 252    ofs-delta
 253        Indicate that the client understands PACKv2 with delta referring
 254        to its base by position in pack rather than by an oid.  That is,
 255        they can read OBJ_OFS_DELTA (ake type 6) in a packfile.
 256
 257If the 'shallow' feature is advertised the following arguments can be
 258included in the clients request as well as the potential addition of the
 259'shallow-info' section in the server's response as explained below.
 260
 261    shallow <oid>
 262        A client must notify the server of all commits for which it only
 263        has shallow copies (meaning that it doesn't have the parents of
 264        a commit) by supplying a 'shallow <oid>' line for each such
 265        object so that the server is aware of the limitations of the
 266        client's history.  This is so that the server is aware that the
 267        client may not have all objects reachable from such commits.
 268
 269    deepen <depth>
 270        Requests that the fetch/clone should be shallow having a commit
 271        depth of <depth> relative to the remote side.
 272
 273    deepen-relative
 274        Requests that the semantics of the "deepen" command be changed
 275        to indicate that the depth requested is relative to the client's
 276        current shallow boundary, instead of relative to the requested
 277        commits.
 278
 279    deepen-since <timestamp>
 280        Requests that the shallow clone/fetch should be cut at a
 281        specific time, instead of depth.  Internally it's equivalent to
 282        doing "git rev-list --max-age=<timestamp>". Cannot be used with
 283        "deepen".
 284
 285    deepen-not <rev>
 286        Requests that the shallow clone/fetch should be cut at a
 287        specific revision specified by '<rev>', instead of a depth.
 288        Internally it's equivalent of doing "git rev-list --not <rev>".
 289        Cannot be used with "deepen", but can be used with
 290        "deepen-since".
 291
 292If the 'filter' feature is advertised, the following argument can be
 293included in the client's request:
 294
 295    filter <filter-spec>
 296        Request that various objects from the packfile be omitted
 297        using one of several filtering techniques. These are intended
 298        for use with partial clone and partial fetch operations. See
 299        `rev-list` for possible "filter-spec" values. When communicating
 300        with other processes, senders SHOULD translate scaled integers
 301        (e.g. "1k") into a fully-expanded form (e.g. "1024") to aid
 302        interoperability with older receivers that may not understand
 303        newly-invented scaling suffixes. However, receivers SHOULD
 304        accept the following suffixes: 'k', 'm', and 'g' for 1024,
 305        1048576, and 1073741824, respectively.
 306
 307If the 'ref-in-want' feature is advertised, the following argument can
 308be included in the client's request as well as the potential addition of
 309the 'wanted-refs' section in the server's response as explained below.
 310
 311    want-ref <ref>
 312        Indicates to the server that the client wants to retrieve a
 313        particular ref, where <ref> is the full name of a ref on the
 314        server.
 315
 316If the 'sideband-all' feature is advertised, the following argument can be
 317included in the client's request:
 318
 319    sideband-all
 320        Instruct the server to send the whole response multiplexed, not just
 321        the packfile section. All non-flush and non-delim PKT-LINE in the
 322        response (not only in the packfile section) will then start with a byte
 323        indicating its sideband (1, 2, or 3), and the server may send "0005\2"
 324        (a PKT-LINE of sideband 2 with no payload) as a keepalive packet.
 325
 326The response of `fetch` is broken into a number of sections separated by
 327delimiter packets (0001), with each section beginning with its section
 328header.
 329
 330    output = *section
 331    section = (acknowledgments | shallow-info | wanted-refs | packfile)
 332              (flush-pkt | delim-pkt)
 333
 334    acknowledgments = PKT-LINE("acknowledgments" LF)
 335                      (nak | *ack)
 336                      (ready)
 337    ready = PKT-LINE("ready" LF)
 338    nak = PKT-LINE("NAK" LF)
 339    ack = PKT-LINE("ACK" SP obj-id LF)
 340
 341    shallow-info = PKT-LINE("shallow-info" LF)
 342                   *PKT-LINE((shallow | unshallow) LF)
 343    shallow = "shallow" SP obj-id
 344    unshallow = "unshallow" SP obj-id
 345
 346    wanted-refs = PKT-LINE("wanted-refs" LF)
 347                  *PKT-LINE(wanted-ref LF)
 348    wanted-ref = obj-id SP refname
 349
 350    packfile = PKT-LINE("packfile" LF)
 351               *PKT-LINE(%x01-03 *%x00-ff)
 352
 353    acknowledgments section
 354        * If the client determines that it is finished with negotiations
 355          by sending a "done" line, the acknowledgments sections MUST be
 356          omitted from the server's response.
 357
 358        * Always begins with the section header "acknowledgments"
 359
 360        * The server will respond with "NAK" if none of the object ids sent
 361          as have lines were common.
 362
 363        * The server will respond with "ACK obj-id" for all of the
 364          object ids sent as have lines which are common.
 365
 366        * A response cannot have both "ACK" lines as well as a "NAK"
 367          line.
 368
 369        * The server will respond with a "ready" line indicating that
 370          the server has found an acceptable common base and is ready to
 371          make and send a packfile (which will be found in the packfile
 372          section of the same response)
 373
 374        * If the server has found a suitable cut point and has decided
 375          to send a "ready" line, then the server can decide to (as an
 376          optimization) omit any "ACK" lines it would have sent during
 377          its response.  This is because the server will have already
 378          determined the objects it plans to send to the client and no
 379          further negotiation is needed.
 380
 381    shallow-info section
 382        * If the client has requested a shallow fetch/clone, a shallow
 383          client requests a fetch or the server is shallow then the
 384          server's response may include a shallow-info section.  The
 385          shallow-info section will be included if (due to one of the
 386          above conditions) the server needs to inform the client of any
 387          shallow boundaries or adjustments to the clients already
 388          existing shallow boundaries.
 389
 390        * Always begins with the section header "shallow-info"
 391
 392        * If a positive depth is requested, the server will compute the
 393          set of commits which are no deeper than the desired depth.
 394
 395        * The server sends a "shallow obj-id" line for each commit whose
 396          parents will not be sent in the following packfile.
 397
 398        * The server sends an "unshallow obj-id" line for each commit
 399          which the client has indicated is shallow, but is no longer
 400          shallow as a result of the fetch (due to its parents being
 401          sent in the following packfile).
 402
 403        * The server MUST NOT send any "unshallow" lines for anything
 404          which the client has not indicated was shallow as a part of
 405          its request.
 406
 407        * This section is only included if a packfile section is also
 408          included in the response.
 409
 410    wanted-refs section
 411        * This section is only included if the client has requested a
 412          ref using a 'want-ref' line and if a packfile section is also
 413          included in the response.
 414
 415        * Always begins with the section header "wanted-refs".
 416
 417        * The server will send a ref listing ("<oid> <refname>") for
 418          each reference requested using 'want-ref' lines.
 419
 420        * The server MUST NOT send any refs which were not requested
 421          using 'want-ref' lines.
 422
 423    packfile section
 424        * This section is only included if the client has sent 'want'
 425          lines in its request and either requested that no more
 426          negotiation be done by sending 'done' or if the server has
 427          decided it has found a sufficient cut point to produce a
 428          packfile.
 429
 430        * Always begins with the section header "packfile"
 431
 432        * The transmission of the packfile begins immediately after the
 433          section header
 434
 435        * The data transfer of the packfile is always multiplexed, using
 436          the same semantics of the 'side-band-64k' capability from
 437          protocol version 1.  This means that each packet, during the
 438          packfile data stream, is made up of a leading 4-byte pkt-line
 439          length (typical of the pkt-line format), followed by a 1-byte
 440          stream code, followed by the actual data.
 441
 442          The stream code can be one of:
 443                1 - pack data
 444                2 - progress messages
 445                3 - fatal error message just before stream aborts
 446
 447server-option
 448~~~~~~~~~~~~~
 449
 450If advertised, indicates that any number of server specific options can be
 451included in a request.  This is done by sending each option as a
 452"server-option=<option>" capability line in the capability-list section of
 453a request.
 454
 455The provided options must not contain a NUL or LF character.