1 Git Protocol Capabilities
2 =========================
4 Servers SHOULD support all capabilities defined in this document.
6 On the very first line of the initial server response of either
7 receive-pack and upload-pack the first reference is followed by
8 a NUL byte and then a list of space delimited server capabilities.
9 These allow the server to declare what it can and cannot support
12 Client will then send a space separated list of capabilities it wants
13 to be in effect. The client MUST NOT ask for capabilities the server
14 did not say it supports.
16 Server MUST diagnose and abort if capabilities it does not understand
17 was sent. Server MUST NOT ignore capabilities that client requested
18 and server advertised. As a consequence of these rules, server MUST
19 NOT advertise capabilities it does not understand.
21 The 'atomic', 'report-status', 'delete-refs', 'quiet', and 'push-cert'
22 capabilities are sent and recognized by the receive-pack (push to server)
25 The 'ofs-delta' and 'side-band-64k' capabilities are sent and recognized
26 by both upload-pack and receive-pack protocols. The 'agent' capability
27 may optionally be sent in both protocols.
29 All other capabilities are only recognized by the upload-pack (fetch
35 The 'multi_ack' capability allows the server to return "ACK obj-id
36 continue" as soon as it finds a commit that it can use as a common
37 base, between the client's wants and the client's have set.
39 By sending this early, the server can potentially head off the client
40 from walking any further down that particular branch of the client's
41 repository history. The client may still need to walk down other
42 branches, sending have lines for those, until the server has a
43 complete cut across the DAG, or the client has said "done".
45 Without multi_ack, a client sends have lines in --date-order until
46 the server has found a common base. That means the client will send
47 have lines that are already known by the server to be common, because
48 they overlap in time with another branch that the server hasn't found
51 For example suppose the client has commits in caps that the server
52 doesn't and the server has commits in lower case that the client
53 doesn't, as in the following diagram:
55 +---- u ---------------------- x
58 a -- b -- c -- d -- E -- F
62 If the client wants x,y and starts out by saying have F,S, the server
63 doesn't know what F,S is. Eventually the client says "have d" and
64 the server sends "ACK d continue" to let the client know to stop
65 walking down that line (so don't send c-b-a), but it's not done yet,
66 it needs a base for x. The client keeps going with S-R-Q, until a
67 gets reached, at which point the server has a clear base and it all
70 Without multi_ack the client would have sent that c-b-a chain anyway,
71 interleaved with S-R-Q.
75 This is an extension of multi_ack that permits client to better
76 understand the server's in-memory state. See pack-protocol.txt,
77 section "Packfile Negotiation" for more information.
81 This capability should only be used with the smart HTTP protocol. If
82 multi_ack_detailed and no-done are both present, then the sender is
83 free to immediately send a pack following its first "ACK obj-id ready"
86 Without no-done in the smart HTTP protocol, the server session would
87 end and the client has to make another trip to send "done" before
88 the server can send the pack. no-done removes the last round and
89 thus slightly reduces latency.
94 A thin pack is one with deltas which reference base objects not
95 contained within the pack (but are known to exist at the receiving
96 end). This can reduce the network traffic significantly, but it
97 requires the receiving end to know how to "thicken" these packs by
98 adding the missing bases to the pack.
100 The upload-pack server advertises 'thin-pack' when it can generate
101 and send a thin pack. A client requests the 'thin-pack' capability
102 when it understands how to "thicken" it, notifying the server that
103 it can receive such a pack. A client MUST NOT request the
104 'thin-pack' capability if it cannot turn a thin pack into a
107 Receive-pack, on the other hand, is assumed by default to be able to
108 handle thin packs, but can ask the client not to use the feature by
109 advertising the 'no-thin' capability. A client MUST NOT send a thin
110 pack if the server advertises the 'no-thin' capability.
112 The reasons for this asymmetry are historical. The receive-pack
113 program did not exist until after the invention of thin packs, so
114 historically the reference implementation of receive-pack always
115 understood thin packs. Adding 'no-thin' later allowed receive-pack
116 to disable the feature in a backwards-compatible manner.
119 side-band, side-band-64k
120 ------------------------
122 This capability means that server can send, and client understand multiplexed
123 progress reports and error info interleaved with the packfile itself.
125 These two options are mutually exclusive. A modern client always
126 favors 'side-band-64k'.
128 Either mode indicates that the packfile data will be streamed broken
129 up into packets of up to either 1000 bytes in the case of 'side_band',
130 or 65520 bytes in the case of 'side_band_64k'. Each packet is made up
131 of a leading 4-byte pkt-line length of how much data is in the packet,
132 followed by a 1-byte stream code, followed by the actual data.
134 The stream code can be one of:
137 2 - progress messages
138 3 - fatal error message just before stream aborts
140 The "side-band-64k" capability came about as a way for newer clients
141 that can handle much larger packets to request packets that are
142 actually crammed nearly full, while maintaining backward compatibility
143 for the older clients.
145 Further, with side-band and its up to 1000-byte messages, it's actually
146 999 bytes of payload and 1 byte for the stream code. With side-band-64k,
147 same deal, you have up to 65519 bytes of data and 1 byte for the stream
150 The client MUST send only maximum of one of "side-band" and "side-
151 band-64k". Server MUST diagnose it as an error if client requests
157 Server can send, and client understand PACKv2 with delta referring to
158 its base by position in pack rather than by an obj-id. That is, they can
159 send/read OBJ_OFS_DELTA (aka type 6) in a packfile.
164 The server may optionally send a capability of the form `agent=X` to
165 notify the client that the server is running version `X`. The client may
166 optionally return its own agent string by responding with an `agent=Y`
167 capability (but it MUST NOT do so if the server did not mention the
168 agent capability). The `X` and `Y` strings may contain any printable
169 ASCII characters except space (i.e., the byte range 32 < x < 127), and
170 are typically of the form "package/version" (e.g., "git/1.8.3.1"). The
171 agent strings are purely informative for statistics and debugging
172 purposes, and MUST NOT be used to programmatically assume the presence
173 or absence of particular features.
178 This capability adds "deepen", "shallow" and "unshallow" commands to
179 the fetch-pack/upload-pack protocol so clients can request shallow
185 This capability adds "deepen-since" command to fetch-pack/upload-pack
186 protocol so the client can request shallow clones that are cut at a
187 specific time, instead of depth. Internally it's equivalent of doing
188 "rev-list --max-age=<timestamp>" on the server side. "deepen-since"
189 cannot be used with "deepen".
194 This capability adds "deepen-not" command to fetch-pack/upload-pack
195 protocol so the client can request shallow clones that are cut at a
196 specific revision, instead of depth. Internally it's equivalent of
197 doing "rev-list --not <rev>" on the server side. "deepen-not"
198 cannot be used with "deepen", but can be used with "deepen-since".
203 If this capability is requested by the client, the semantics of
204 "deepen" command is changed. The "depth" argument is the depth from
205 the current shallow boundary, instead of the depth from remote refs.
210 The client was started with "git clone -q" or something, and doesn't
211 want that side band 2. Basically the client just says "I do not
212 wish to receive stream 2 on sideband, so do not send it to me, and if
213 you did, I will drop it on the floor anyway". However, the sideband
214 channel 3 is still used for error responses.
219 The 'include-tag' capability is about sending annotated tags if we are
220 sending objects they point to. If we pack an object to the client, and
221 a tag object points exactly at that object, we pack the tag object too.
222 In general this allows a client to get all new annotated tags when it
223 fetches a branch, in a single network connection.
225 Clients MAY always send include-tag, hardcoding it into a request when
226 the server advertises this capability. The decision for a client to
227 request include-tag only has to do with the client's desires for tag
228 data, whether or not a server had advertised objects in the
229 refs/tags/* namespace.
231 Servers MUST pack the tags if their referrant is packed and the client
232 has requested include-tags.
234 Clients MUST be prepared for the case where a server has ignored
235 include-tag and has not actually sent tags in the pack. In such
236 cases the client SHOULD issue a subsequent fetch to acquire the tags
237 that include-tag would have otherwise given the client.
239 The server SHOULD send include-tag, if it supports it, regardless
240 of whether or not there are tags available.
245 The receive-pack process can receive a 'report-status' capability,
246 which tells it that the client wants a report of what happened after
247 a packfile upload and reference update. If the pushing client requests
248 this capability, after unpacking and updating references the server
249 will respond with whether the packfile unpacked successfully and if
250 each reference was updated successfully. If any of those were not
251 successful, it will send back an error message. See pack-protocol.txt
252 for example messages.
257 If the server sends back the 'delete-refs' capability, it means that
258 it is capable of accepting a zero-id value as the target
259 value of a reference update. It is not sent back by the client, it
260 simply informs the client that it can be sent zero-id values
261 to delete references.
266 If the receive-pack server advertises the 'quiet' capability, it is
267 capable of silencing human-readable progress output which otherwise may
268 be shown when processing the received pack. A send-pack client should
269 respond with the 'quiet' capability to suppress server-side progress
270 reporting if the local progress reporting is also being suppressed
271 (e.g., via `push -q`, or if stderr does not go to a tty).
276 If the server sends the 'atomic' capability it is capable of accepting
277 atomic pushes. If the pushing client requests this capability, the server
278 will update the refs in one atomic transaction. Either all refs are
284 If the server sends the 'push-options' capability it is able to accept
285 push options after the update commands have been sent, but before the
286 packfile is streamed. If the pushing client requests this capability,
287 the server will pass the options to the pre- and post- receive hooks
288 that process this push request.
290 allow-tip-sha1-in-want
291 ----------------------
293 If the upload-pack server advertises this capability, fetch-pack may
294 send "want" lines with SHA-1s that exist at the server but are not
295 advertised by upload-pack.
297 allow-reachable-sha1-in-want
298 ----------------------------
300 If the upload-pack server advertises this capability, fetch-pack may
301 send "want" lines with SHA-1s that exist at the server but are not
302 advertised by upload-pack.
307 The receive-pack server that advertises this capability is willing
308 to accept a signed push certificate, and asks the <nonce> to be
309 included in the push certificate. A send-pack client MUST NOT
310 send a push-cert packet unless the receive-pack server advertises
projects / git / blob