1Git Protocol Capabilities 2========================= 3 4Servers SHOULD support all capabilities defined in this document. 5 6On the very first line of the initial server response of either 7receive-pack and upload-pack the first reference is followed by 8a NUL byte and then a list of space delimited server capabilities. 9These allow the server to declare what it can and cannot support 10to the client. 11 12Client will then send a space separated list of capabilities it wants 13to be in effect. The client MUST NOT ask for capabilities the server 14did not say it supports. 15 16Server MUST diagnose and abort if capabilities it does not understand 17was sent. Server MUST NOT ignore capabilities that client requested 18and server advertised. As a consequence of these rules, server MUST 19NOT advertise capabilities it does not understand. 20 21The 'report-status', 'delete-refs', 'quiet', and 'push-cert' capabilities 22are sent and recognized by the receive-pack (push to server) process. 23 24The 'ofs-delta' and 'side-band-64k' capabilities are sent and recognized 25by both upload-pack and receive-pack protocols. The 'agent' capability 26may optionally be sent in both protocols. 27 28All other capabilities are only recognized by the upload-pack (fetch 29from server) process. 30 31multi_ack 32--------- 33 34The 'multi_ack' capability allows the server to return "ACK obj-id 35continue" as soon as it finds a commit that it can use as a common 36base, between the client's wants and the client's have set. 37 38By sending this early, the server can potentially head off the client 39from walking any further down that particular branch of the client's 40repository history. The client may still need to walk down other 41branches, sending have lines for those, until the server has a 42complete cut across the DAG, or the client has said "done". 43 44Without multi_ack, a client sends have lines in --date-order until 45the server has found a common base. That means the client will send 46have lines that are already known by the server to be common, because 47they overlap in time with another branch that the server hasn't found 48a common base on yet. 49 50For example suppose the client has commits in caps that the server 51doesn't and the server has commits in lower case that the client 52doesn't, as in the following diagram: 53 54 +---- u ---------------------- x 55 / +----- y 56 / / 57 a -- b -- c -- d -- E -- F 58 \ 59 +--- Q -- R -- S 60 61If the client wants x,y and starts out by saying have F,S, the server 62doesn't know what F,S is. Eventually the client says "have d" and 63the server sends "ACK d continue" to let the client know to stop 64walking down that line (so don't send c-b-a), but it's not done yet, 65it needs a base for x. The client keeps going with S-R-Q, until a 66gets reached, at which point the server has a clear base and it all 67ends. 68 69Without multi_ack the client would have sent that c-b-a chain anyway, 70interleaved with S-R-Q. 71 72multi_ack_detailed 73------------------ 74This is an extension of multi_ack that permits client to better 75understand the server's in-memory state. See pack-protocol.txt, 76section "Packfile Negotiation" for more information. 77 78no-done 79------- 80This capability should only be used with the smart HTTP protocol. If 81multi_ack_detailed and no-done are both present, then the sender is 82free to immediately send a pack following its first "ACK obj-id ready" 83message. 84 85Without no-done in the smart HTTP protocol, the server session would 86end and the client has to make another trip to send "done" before 87the server can send the pack. no-done removes the last round and 88thus slightly reduces latency. 89 90thin-pack 91--------- 92 93A thin pack is one with deltas which reference base objects not 94contained within the pack (but are known to exist at the receiving 95end). This can reduce the network traffic significantly, but it 96requires the receiving end to know how to "thicken" these packs by 97adding the missing bases to the pack. 98 99The upload-pack server advertises 'thin-pack' when it can generate 100and send a thin pack. A client requests the 'thin-pack' capability 101when it understands how to "thicken" it, notifying the server that 102it can receive such a pack. A client MUST NOT request the 103'thin-pack' capability if it cannot turn a thin pack into a 104self-contained pack. 105 106Receive-pack, on the other hand, is assumed by default to be able to 107handle thin packs, but can ask the client not to use the feature by 108advertising the 'no-thin' capability. A client MUST NOT send a thin 109pack if the server advertises the 'no-thin' capability. 110 111The reasons for this asymmetry are historical. The receive-pack 112program did not exist until after the invention of thin packs, so 113historically the reference implementation of receive-pack always 114understood thin packs. Adding 'no-thin' later allowed receive-pack 115to disable the feature in a backwards-compatible manner. 116 117 118side-band, side-band-64k 119------------------------ 120 121This capability means that server can send, and client understand multiplexed 122progress reports and error info interleaved with the packfile itself. 123 124These two options are mutually exclusive. A modern client always 125favors 'side-band-64k'. 126 127Either mode indicates that the packfile data will be streamed broken 128up into packets of up to either 1000 bytes in the case of 'side_band', 129or 65520 bytes in the case of 'side_band_64k'. Each packet is made up 130of a leading 4-byte pkt-line length of how much data is in the packet, 131followed by a 1-byte stream code, followed by the actual data. 132 133The stream code can be one of: 134 135 1 - pack data 136 2 - progress messages 137 3 - fatal error message just before stream aborts 138 139The "side-band-64k" capability came about as a way for newer clients 140that can handle much larger packets to request packets that are 141actually crammed nearly full, while maintaining backward compatibility 142for the older clients. 143 144Further, with side-band and its up to 1000-byte messages, it's actually 145999 bytes of payload and 1 byte for the stream code. With side-band-64k, 146same deal, you have up to 65519 bytes of data and 1 byte for the stream 147code. 148 149The client MUST send only maximum of one of "side-band" and "side- 150band-64k". Server MUST diagnose it as an error if client requests 151both. 152 153ofs-delta 154--------- 155 156Server can send, and client understand PACKv2 with delta referring to 157its base by position in pack rather than by an obj-id. That is, they can 158send/read OBJ_OFS_DELTA (aka type 6) in a packfile. 159 160agent 161----- 162 163The server may optionally send a capability of the form `agent=X` to 164notify the client that the server is running version `X`. The client may 165optionally return its own agent string by responding with an `agent=Y` 166capability (but it MUST NOT do so if the server did not mention the 167agent capability). The `X` and `Y` strings may contain any printable 168ASCII characters except space (i.e., the byte range 32 < x < 127), and 169are typically of the form "package/version" (e.g., "git/1.8.3.1"). The 170agent strings are purely informative for statistics and debugging 171purposes, and MUST NOT be used to programatically assume the presence 172or absence of particular features. 173 174shallow 175------- 176 177This capability adds "deepen", "shallow" and "unshallow" commands to 178the fetch-pack/upload-pack protocol so clients can request shallow 179clones. 180 181no-progress 182----------- 183 184The client was started with "git clone -q" or something, and doesn't 185want that side band 2. Basically the client just says "I do not 186wish to receive stream 2 on sideband, so do not send it to me, and if 187you did, I will drop it on the floor anyway". However, the sideband 188channel 3 is still used for error responses. 189 190include-tag 191----------- 192 193The 'include-tag' capability is about sending annotated tags if we are 194sending objects they point to. If we pack an object to the client, and 195a tag object points exactly at that object, we pack the tag object too. 196In general this allows a client to get all new annotated tags when it 197fetches a branch, in a single network connection. 198 199Clients MAY always send include-tag, hardcoding it into a request when 200the server advertises this capability. The decision for a client to 201request include-tag only has to do with the client's desires for tag 202data, whether or not a server had advertised objects in the 203refs/tags/* namespace. 204 205Servers MUST pack the tags if their referrant is packed and the client 206has requested include-tags. 207 208Clients MUST be prepared for the case where a server has ignored 209include-tag and has not actually sent tags in the pack. In such 210cases the client SHOULD issue a subsequent fetch to acquire the tags 211that include-tag would have otherwise given the client. 212 213The server SHOULD send include-tag, if it supports it, regardless 214of whether or not there are tags available. 215 216report-status 217------------- 218 219The receive-pack process can receive a 'report-status' capability, 220which tells it that the client wants a report of what happened after 221a packfile upload and reference update. If the pushing client requests 222this capability, after unpacking and updating references the server 223will respond with whether the packfile unpacked successfully and if 224each reference was updated successfully. If any of those were not 225successful, it will send back an error message. See pack-protocol.txt 226for example messages. 227 228delete-refs 229----------- 230 231If the server sends back the 'delete-refs' capability, it means that 232it is capable of accepting a zero-id value as the target 233value of a reference update. It is not sent back by the client, it 234simply informs the client that it can be sent zero-id values 235to delete references. 236 237quiet 238----- 239 240If the receive-pack server advertises the 'quiet' capability, it is 241capable of silencing human-readable progress output which otherwise may 242be shown when processing the received pack. A send-pack client should 243respond with the 'quiet' capability to suppress server-side progress 244reporting if the local progress reporting is also being suppressed 245(e.g., via `push -q`, or if stderr does not go to a tty). 246 247allow-tip-sha1-in-want 248---------------------- 249 250If the upload-pack server advertises this capability, fetch-pack may 251send "want" lines with SHA-1s that exist at the server but are not 252advertised by upload-pack. 253 254push-cert=<nonce> 255----------------- 256 257The receive-pack server that advertises this capability is willing 258to accept a signed push certificate, and asks the <nonce> to be 259included in the push certificate. A send-pack client MUST NOT 260send a push-cert packet unless the receive-pack server advertises 261this capability.