1 Git Wire Protocol, Version 2
2 ============================
4 This document presents a specification for a version 2 of Git's wire
5 protocol. Protocol v2 will improve upon v1 in the following ways:
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
19 In protocol v2 communication is command oriented. When first contacting a
20 server a list of capabilities will advertised. Some of these capabilities
21 will be commands which a client can request be executed. Once a command
22 has completed, a client can reuse the connection and request that other
28 All 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.
32 In protocol v2 these special packets will have the following semantics:
34 * '0000' Flush Packet (flush-pkt) - indicates the end of a message
35 * '0001' Delimiter Packet (delim-pkt) - separates sections of a message
36 * '0002' Response End Packet (response-end-pkt) - indicates the end of a
37 response for stateless connections
39 Initial Client Request
40 ----------------------
42 In general a client can request to speak protocol v2 by sending
43 `version=2` through the respective side-channel for the transport being
44 used which inevitably sets `GIT_PROTOCOL`. More information can be
45 found in `pack-protocol.txt` and `http-protocol.txt`, as well as the
46 `GIT_PROTOCOL` definition in `git.txt`. In all cases the
47 response from the server is the capability advertisement.
52 When using the git:// transport, you can request to use protocol v2 by
53 sending "version=2" as an extra parameter:
55 003egit-upload-pack /project.git\0host=myserver.com\0\0version=2\0
57 SSH and File Transport
58 ~~~~~~~~~~~~~~~~~~~~~~
60 When using either the ssh:// or file:// transport, the GIT_PROTOCOL
61 environment variable must be set explicitly to include "version=2".
62 The server may need to be configured to allow this environment variable
68 When using the http:// or https:// transport a client makes a "smart"
69 info/refs request as described in `http-protocol.txt` and requests that
70 v2 be used by supplying "version=2" in the `Git-Protocol` header.
72 C: GET $GIT_URL/info/refs?service=git-upload-pack HTTP/1.0
73 C: Git-Protocol: version=2
75 A v2 server would reply:
82 S: <capability-advertisement>
84 Subsequent requests are then made directly to the service
85 `$GIT_URL/git-upload-pack`. (This works the same for git-receive-pack).
87 Uses the `--http-backend-info-refs` option to
88 linkgit:git-upload-pack[1].
90 The server may need to be configured to pass this header's contents via
91 the `GIT_PROTOCOL` variable. See the discussion in `git-http-backend.txt`.
93 Capability Advertisement
94 ------------------------
96 A server which decides to communicate (based on a request from a client)
97 using protocol version 2, notifies the client by sending a version string
98 in its initial response followed by an advertisement of its capabilities.
99 Each capability is a key with an optional value. Clients must ignore all
100 unknown keys. Semantics of unknown values are left to the definition of
101 each key. Some capabilities will describe commands which can be requested
102 to be executed by the client.
104 capability-advertisement = protocol-version
108 protocol-version = PKT-LINE("version 2" LF)
109 capability-list = *capability
110 capability = PKT-LINE(key[=value] LF)
112 key = 1*(ALPHA | DIGIT | "-_")
113 value = 1*(ALPHA | DIGIT | " -_.,?\/{}[]()<>!@#$%^&*+=:;")
118 After receiving the capability advertisement, a client can then issue a
119 request to select the command it wants with any particular capabilities
120 or arguments. There is then an optional section where the client can
121 provide any command specific parameters or queries. Only a single
122 command can be requested at a time.
124 request = empty-request | command-request
125 empty-request = flush-pkt
126 command-request = command
131 command = PKT-LINE("command=" key LF)
132 command-args = *command-specific-arg
134 command-specific-args are packet line framed arguments defined by
135 each individual command.
137 The server will then check to ensure that the client's request is
138 comprised of a valid command as well as valid capabilities which were
139 advertised. If the request is valid the server will then execute the
140 command. A server MUST wait till it has received the client's entire
141 request before issuing a response. The format of the response is
142 determined by the command being executed, but in all cases a flush-pkt
143 indicates the end of the response.
145 When a command has finished, and the client has received the entire
146 response from the server, a client can either request that another
147 command be executed or can terminate the connection. A client may
148 optionally send an empty request consisting of just a flush-pkt to
149 indicate that no more requests will be made.
154 There are two different types of capabilities: normal capabilities,
155 which can be used to convey information or alter the behavior of a
156 request, and commands, which are the core actions that a client wants to
157 perform (fetch, push, etc).
159 Protocol version 2 is stateless by default. This means that all commands
160 must only last a single round and be stateless from the perspective of the
161 server side, unless the client has requested a capability indicating that
162 state should be maintained by the server. Clients MUST NOT require state
163 management on the server side in order to function correctly. This
164 permits simple round-robin load-balancing on the server side, without
165 needing to worry about state management.
170 The server can advertise the `agent` capability with a value `X` (in the
171 form `agent=X`) to notify the client that the server is running version
172 `X`. The client may optionally send its own agent string by including
173 the `agent` capability with a value `Y` (in the form `agent=Y`) in its
174 request to the server (but it MUST NOT do so if the server did not
175 advertise the agent capability). The `X` and `Y` strings may contain any
176 printable ASCII characters except space (i.e., the byte range 32 < x <
177 127), and are typically of the form "package/version" (e.g.,
178 "git/1.8.3.1"). The agent strings are purely informative for statistics
179 and debugging purposes, and MUST NOT be used to programmatically assume
180 the presence or absence of particular features.
185 `ls-refs` is the command used to request a reference advertisement in v2.
186 Unlike the current reference advertisement, ls-refs takes in arguments
187 which can be used to limit the refs sent from the server.
189 Additional features not supported in the base command will be advertised
190 as the value of the command in the capability advertisement in the form
191 of a space separated list of features: "<command>=<feature 1> <feature 2>"
193 ls-refs takes in the following arguments:
196 In addition to the object pointed by it, show the underlying ref
197 pointed by it when showing a symbolic ref.
201 When specified, only references having a prefix matching one of
202 the provided prefixes are displayed. Multiple instances may be
203 given, in which case references matching any prefix will be
204 shown. Note that this is purely for optimization; a server MAY
205 show refs not matching the prefix if it chooses, and clients
206 should filter the result themselves.
208 If the 'unborn' feature is advertised the following argument can be
209 included in the client's request.
212 The server will send information about HEAD even if it is a symref
213 pointing to an unborn branch in the form "unborn HEAD
214 symref-target:<target>".
216 The output of ls-refs is as follows:
220 obj-id-or-unborn = (obj-id | "unborn")
221 ref = PKT-LINE(obj-id-or-unborn SP refname *(SP ref-attribute) LF)
222 ref-attribute = (symref | peeled)
223 symref = "symref-target:" symref-target
224 peeled = "peeled:" obj-id
229 `fetch` is the command used to fetch a packfile in v2. It can be looked
230 at as a modified version of the v1 fetch where the ref-advertisement is
231 stripped out (since the `ls-refs` command fills that role) and the
232 message format is tweaked to eliminate redundancies and permit easy
233 addition of future extensions.
235 Additional features not supported in the base command will be advertised
236 as the value of the command in the capability advertisement in the form
237 of a space separated list of features: "<command>=<feature 1> <feature 2>"
239 A `fetch` request can take the following arguments:
242 Indicates to the server an object which the client wants to
243 retrieve. Wants can be anything and are not limited to
247 Indicates to the server an object which the client has locally.
248 This allows the server to make a packfile which only contains
249 the objects that the client needs. Multiple 'have' lines can be
253 Indicates to the server that negotiation should terminate (or
254 not even begin if performing a clone) and that the server should
255 use the information supplied in the request to construct the
259 Request that a thin pack be sent, which is a pack with deltas
260 which reference base objects not contained within the pack (but
261 are known to exist at the receiving end). This can reduce the
262 network traffic significantly, but it requires the receiving end
263 to know how to "thicken" these packs by adding the missing bases
267 Request that progress information that would normally be sent on
268 side-band channel 2, during the packfile transfer, should not be
269 sent. However, the side-band channel 3 is still used for error
273 Request that annotated tags should be sent if the objects they
274 point to are being sent.
277 Indicate that the client understands PACKv2 with delta referring
278 to its base by position in pack rather than by an oid. That is,
279 they can read OBJ_OFS_DELTA (aka type 6) in a packfile.
281 If the 'shallow' feature is advertised the following arguments can be
282 included in the clients request as well as the potential addition of the
283 'shallow-info' section in the server's response as explained below.
286 A client must notify the server of all commits for which it only
287 has shallow copies (meaning that it doesn't have the parents of
288 a commit) by supplying a 'shallow <oid>' line for each such
289 object so that the server is aware of the limitations of the
290 client's history. This is so that the server is aware that the
291 client may not have all objects reachable from such commits.
294 Requests that the fetch/clone should be shallow having a commit
295 depth of <depth> relative to the remote side.
298 Requests that the semantics of the "deepen" command be changed
299 to indicate that the depth requested is relative to the client's
300 current shallow boundary, instead of relative to the requested
303 deepen-since <timestamp>
304 Requests that the shallow clone/fetch should be cut at a
305 specific time, instead of depth. Internally it's equivalent to
306 doing "git rev-list --max-age=<timestamp>". Cannot be used with
310 Requests that the shallow clone/fetch should be cut at a
311 specific revision specified by '<rev>', instead of a depth.
312 Internally it's equivalent of doing "git rev-list --not <rev>".
313 Cannot be used with "deepen", but can be used with
316 If the 'filter' feature is advertised, the following argument can be
317 included in the client's request:
320 Request that various objects from the packfile be omitted
321 using one of several filtering techniques. These are intended
322 for use with partial clone and partial fetch operations. See
323 `rev-list` for possible "filter-spec" values. When communicating
324 with other processes, senders SHOULD translate scaled integers
325 (e.g. "1k") into a fully-expanded form (e.g. "1024") to aid
326 interoperability with older receivers that may not understand
327 newly-invented scaling suffixes. However, receivers SHOULD
328 accept the following suffixes: 'k', 'm', and 'g' for 1024,
329 1048576, and 1073741824, respectively.
331 If the 'ref-in-want' feature is advertised, the following argument can
332 be included in the client's request as well as the potential addition of
333 the 'wanted-refs' section in the server's response as explained below.
336 Indicates to the server that the client wants to retrieve a
337 particular ref, where <ref> is the full name of a ref on the
340 If the 'sideband-all' feature is advertised, the following argument can be
341 included in the client's request:
344 Instruct the server to send the whole response multiplexed, not just
345 the packfile section. All non-flush and non-delim PKT-LINE in the
346 response (not only in the packfile section) will then start with a byte
347 indicating its sideband (1, 2, or 3), and the server may send "0005\2"
348 (a PKT-LINE of sideband 2 with no payload) as a keepalive packet.
350 If the 'packfile-uris' feature is advertised, the following argument
351 can be included in the client's request as well as the potential
352 addition of the 'packfile-uris' section in the server's response as
355 packfile-uris <comma-separated list of protocols>
356 Indicates to the server that the client is willing to receive
357 URIs of any of the given protocols in place of objects in the
358 sent packfile. Before performing the connectivity check, the
359 client should download from all given URIs. Currently, the
360 protocols supported are "http" and "https".
362 If the 'wait-for-done' feature is advertised, the following argument
363 can be included in the client's request.
366 Indicates to the server that it should never send "ready", but
367 should wait for the client to say "done" before sending the
370 The response of `fetch` is broken into a number of sections separated by
371 delimiter packets (0001), with each section beginning with its section
372 header. Most sections are sent only when the packfile is sent.
374 output = acknowledgements flush-pkt |
375 [acknowledgments delim-pkt] [shallow-info delim-pkt]
376 [wanted-refs delim-pkt] [packfile-uris delim-pkt]
379 acknowledgments = PKT-LINE("acknowledgments" LF)
382 ready = PKT-LINE("ready" LF)
383 nak = PKT-LINE("NAK" LF)
384 ack = PKT-LINE("ACK" SP obj-id LF)
386 shallow-info = PKT-LINE("shallow-info" LF)
387 *PKT-LINE((shallow | unshallow) LF)
388 shallow = "shallow" SP obj-id
389 unshallow = "unshallow" SP obj-id
391 wanted-refs = PKT-LINE("wanted-refs" LF)
392 *PKT-LINE(wanted-ref LF)
393 wanted-ref = obj-id SP refname
395 packfile-uris = PKT-LINE("packfile-uris" LF) *packfile-uri
396 packfile-uri = PKT-LINE(40*(HEXDIGIT) SP *%x20-ff LF)
398 packfile = PKT-LINE("packfile" LF)
399 *PKT-LINE(%x01-03 *%x00-ff)
401 acknowledgments section
402 * If the client determines that it is finished with negotiations by
403 sending a "done" line (thus requiring the server to send a packfile),
404 the acknowledgments sections MUST be omitted from the server's
407 * Always begins with the section header "acknowledgments"
409 * The server will respond with "NAK" if none of the object ids sent
410 as have lines were common.
412 * The server will respond with "ACK obj-id" for all of the
413 object ids sent as have lines which are common.
415 * A response cannot have both "ACK" lines as well as a "NAK"
418 * The server will respond with a "ready" line indicating that
419 the server has found an acceptable common base and is ready to
420 make and send a packfile (which will be found in the packfile
421 section of the same response)
423 * If the server has found a suitable cut point and has decided
424 to send a "ready" line, then the server can decide to (as an
425 optimization) omit any "ACK" lines it would have sent during
426 its response. This is because the server will have already
427 determined the objects it plans to send to the client and no
428 further negotiation is needed.
431 * If the client has requested a shallow fetch/clone, a shallow
432 client requests a fetch or the server is shallow then the
433 server's response may include a shallow-info section. The
434 shallow-info section will be included if (due to one of the
435 above conditions) the server needs to inform the client of any
436 shallow boundaries or adjustments to the clients already
437 existing shallow boundaries.
439 * Always begins with the section header "shallow-info"
441 * If a positive depth is requested, the server will compute the
442 set of commits which are no deeper than the desired depth.
444 * The server sends a "shallow obj-id" line for each commit whose
445 parents will not be sent in the following packfile.
447 * The server sends an "unshallow obj-id" line for each commit
448 which the client has indicated is shallow, but is no longer
449 shallow as a result of the fetch (due to its parents being
450 sent in the following packfile).
452 * The server MUST NOT send any "unshallow" lines for anything
453 which the client has not indicated was shallow as a part of
457 * This section is only included if the client has requested a
458 ref using a 'want-ref' line and if a packfile section is also
459 included in the response.
461 * Always begins with the section header "wanted-refs".
463 * The server will send a ref listing ("<oid> <refname>") for
464 each reference requested using 'want-ref' lines.
466 * The server MUST NOT send any refs which were not requested
467 using 'want-ref' lines.
469 packfile-uris section
470 * This section is only included if the client sent
471 'packfile-uris' and the server has at least one such URI to
474 * Always begins with the section header "packfile-uris".
476 * For each URI the server sends, it sends a hash of the pack's
477 contents (as output by git index-pack) followed by the URI.
479 * The hashes are 40 hex characters long. When Git upgrades to a new
480 hash algorithm, this might need to be updated. (It should match
481 whatever index-pack outputs after "pack\t" or "keep\t".
484 * This section is only included if the client has sent 'want'
485 lines in its request and either requested that no more
486 negotiation be done by sending 'done' or if the server has
487 decided it has found a sufficient cut point to produce a
490 * Always begins with the section header "packfile"
492 * The transmission of the packfile begins immediately after the
495 * The data transfer of the packfile is always multiplexed, using
496 the same semantics of the 'side-band-64k' capability from
497 protocol version 1. This means that each packet, during the
498 packfile data stream, is made up of a leading 4-byte pkt-line
499 length (typical of the pkt-line format), followed by a 1-byte
500 stream code, followed by the actual data.
502 The stream code can be one of:
504 2 - progress messages
505 3 - fatal error message just before stream aborts
510 If advertised, indicates that any number of server specific options can be
511 included in a request. This is done by sending each option as a
512 "server-option=<option>" capability line in the capability-list section of
515 The provided options must not contain a NUL or LF character.
520 The server can advertise the `object-format` capability with a value `X` (in the
521 form `object-format=X`) to notify the client that the server is able to deal
522 with objects using hash algorithm X. If not specified, the server is assumed to
523 only handle SHA-1. If the client would like to use a hash algorithm other than
524 SHA-1, it should specify its object-format string.
526 session-id=<session id>
527 ~~~~~~~~~~~~~~~~~~~~~~~
529 The server may advertise a session ID that can be used to identify this process
530 across multiple requests. The client may advertise its own session ID back to
533 Session IDs should be unique to a given process. They must fit within a
534 packet-line, and must not contain non-printable or whitespace characters. The
535 current implementation uses trace2 session IDs (see
536 link:api-trace2.html[api-trace2] for details), but this may change and users of
537 the session ID should not rely on this fact.
542 `object-info` is the command to retrieve information about one or more objects.
543 Its main purpose is to allow a client to make decisions based on this
544 information without having to fully fetch objects. Object size is the only
545 information that is currently supported.
547 An `object-info` request takes the following arguments:
550 Requests size information to be returned for each listed object id.
553 Indicates to the server an object which the client wants to obtain
556 The response of `object-info` is a list of the requested object ids
557 and associated requested information, each separated by a single space.
559 output = info flush-pkt
561 info = PKT-LINE(attrs) LF)
562 *PKT-LINE(obj-info LF)
564 attrs = attr | attrs SP attrs
568 obj-info = obj-id SP obj-size