Documentation/technical/http-protocol.txt

   1 HTTP transfer protocols
   2 =======================
   3
   4 Git supports two HTTP based transfer protocols.  A "dumb" protocol
   5 which requires only a standard HTTP server on the server end of the
   6 connection, and a "smart" protocol which requires a Git aware CGI
   7 (or server module).  This document describes both protocols.
   8
   9 As a design feature smart clients can automatically upgrade "dumb"
  10 protocol URLs to smart URLs.  This permits all users to have the
  11 same published URL, and the peers automatically select the most
  12 efficient transport available to them.
  13
  14
  15 URL Format
  16 ----------
  17
  18 URLs for Git repositories accessed by HTTP use the standard HTTP
  19 URL syntax documented by RFC 1738, so they are of the form:
  20
  21   http://<host>:<port>/<path>?<searchpart>
  22
  23 Within this documentation the placeholder `$GIT_URL` will stand for
  24 the http:// repository URL entered by the end-user.
  25
  26 Servers SHOULD handle all requests to locations matching `$GIT_URL`, as
  27 both the "smart" and "dumb" HTTP protocols used by Git operate
  28 by appending additional path components onto the end of the user
  29 supplied `$GIT_URL` string.
  30
  31 An example of a dumb client requesting for a loose object:
  32
  33   $GIT_URL:     http://example.com:8080/git/repo.git
  34   URL request:  http://example.com:8080/git/repo.git/objects/d0/49f6c27a2244e12041955e262a404c7faba355
  35
  36 An example of a smart request to a catch-all gateway:
  37
  38   $GIT_URL:     http://example.com/daemon.cgi?svc=git&q=
  39   URL request:  http://example.com/daemon.cgi?svc=git&q=/info/refs&service=git-receive-pack
  40
  41 An example of a request to a submodule:
  42
  43   $GIT_URL:     http://example.com/git/repo.git/path/submodule.git
  44   URL request:  http://example.com/git/repo.git/path/submodule.git/info/refs
  45
  46 Clients MUST strip a trailing `/`, if present, from the user supplied
  47 `$GIT_URL` string to prevent empty path tokens (`//`) from appearing
  48 in any URL sent to a server.  Compatible clients MUST expand
  49 `$GIT_URL/info/refs` as `foo/info/refs` and not `foo//info/refs`.
  50
  51
  52 Authentication
  53 --------------
  54
  55 Standard HTTP authentication is used if authentication is required
  56 to access a repository, and MAY be configured and enforced by the
  57 HTTP server software.
  58
  59 Because Git repositories are accessed by standard path components
  60 server administrators MAY use directory based permissions within
  61 their HTTP server to control repository access.
  62
  63 Clients SHOULD support Basic authentication as described by RFC 2617.
  64 Servers SHOULD support Basic authentication by relying upon the
  65 HTTP server placed in front of the Git server software.
  66
  67 Servers SHOULD NOT require HTTP cookies for the purposes of
  68 authentication or access control.
  69
  70 Clients and servers MAY support other common forms of HTTP based
  71 authentication, such as Digest authentication.
  72
  73
  74 SSL
  75 ---
  76
  77 Clients and servers SHOULD support SSL, particularly to protect
  78 passwords when relying on Basic HTTP authentication.
  79
  80
  81 Session State
  82 -------------
  83
  84 The Git over HTTP protocol (much like HTTP itself) is stateless
  85 from the perspective of the HTTP server side.  All state MUST be
  86 retained and managed by the client process.  This permits simple
  87 round-robin load-balancing on the server side, without needing to
  88 worry about state management.
  89
  90 Clients MUST NOT require state management on the server side in
  91 order to function correctly.
  92
  93 Servers MUST NOT require HTTP cookies in order to function correctly.
  94 Clients MAY store and forward HTTP cookies during request processing
  95 as described by RFC 2616 (HTTP/1.1).  Servers SHOULD ignore any
  96 cookies sent by a client.
  97
  98
  99 General Request Processing
 100 --------------------------
 101
 102 Except where noted, all standard HTTP behavior SHOULD be assumed
 103 by both client and server.  This includes (but is not necessarily
 104 limited to):
 105
 106 If there is no repository at `$GIT_URL`, or the resource pointed to by a
 107 location matching `$GIT_URL` does not exist, the server MUST NOT respond
 108 with `200 OK` response.  A server SHOULD respond with
 109 `404 Not Found`, `410 Gone`, or any other suitable HTTP status code
 110 which does not imply the resource exists as requested.
 111
 112 If there is a repository at `$GIT_URL`, but access is not currently
 113 permitted, the server MUST respond with the `403 Forbidden` HTTP
 114 status code.
 115
 116 Servers SHOULD support both HTTP 1.0 and HTTP 1.1.
 117 Servers SHOULD support chunked encoding for both request and response
 118 bodies.
 119
 120 Clients SHOULD support both HTTP 1.0 and HTTP 1.1.
 121 Clients SHOULD support chunked encoding for both request and response
 122 bodies.
 123
 124 Servers MAY return ETag and/or Last-Modified headers.
 125
 126 Clients MAY revalidate cached entities by including If-Modified-Since
 127 and/or If-None-Match request headers.
 128
 129 Servers MAY return `304 Not Modified` if the relevant headers appear
 130 in the request and the entity has not changed.  Clients MUST treat
 131 `304 Not Modified` identical to `200 OK` by reusing the cached entity.
 132
 133 Clients MAY reuse a cached entity without revalidation if the
 134 Cache-Control and/or Expires header permits caching.  Clients and
 135 servers MUST follow RFC 2616 for cache controls.
 136
 137
 138 Discovering References
 139 ----------------------
 140
 141 All HTTP clients MUST begin either a fetch or a push exchange by
 142 discovering the references available on the remote repository.
 143
 144 Dumb Clients
 145 ~~~~~~~~~~~~
 146
 147 HTTP clients that only support the "dumb" protocol MUST discover
 148 references by making a request for the special info/refs file of
 149 the repository.
 150
 151 Dumb HTTP clients MUST make a `GET` request to `$GIT_URL/info/refs`,
 152 without any search/query parameters.
 153
 154    C: GET $GIT_URL/info/refs HTTP/1.0
 155
 156    S: 200 OK
 157    S:
 158    S: 95dcfa3633004da0049d3d0fa03f80589cbcaf31  refs/heads/maint
 159    S: d049f6c27a2244e12041955e262a404c7faba355  refs/heads/master
 160    S: 2cb58b79488a98d2721cea644875a8dd0026b115  refs/tags/v1.0
 161    S: a3c2e2402b99163d1d59756e5f207ae21cccba4c  refs/tags/v1.0^{}
 162
 163 The Content-Type of the returned info/refs entity SHOULD be
 164 `text/plain; charset=utf-8`, but MAY be any content type.
 165 Clients MUST NOT attempt to validate the returned Content-Type.
 166 Dumb servers MUST NOT return a return type starting with
 167 `application/x-git-`.
 168
 169 Cache-Control headers MAY be returned to disable caching of the
 170 returned entity.
 171
 172 When examining the response clients SHOULD only examine the HTTP
 173 status code.  Valid responses are `200 OK`, or `304 Not Modified`.
 174
 175 The returned content is a UNIX formatted text file describing
 176 each ref and its known value.  The file SHOULD be sorted by name
 177 according to the C locale ordering.  The file SHOULD NOT include
 178 the default ref named `HEAD`.
 179
 180   info_refs   =  *( ref_record )
 181   ref_record  =  any_ref / peeled_ref
 182
 183   any_ref     =  obj-id HTAB refname LF
 184   peeled_ref  =  obj-id HTAB refname LF
 185                  obj-id HTAB refname "^{}" LF
 186
 187 Smart Clients
 188 ~~~~~~~~~~~~~
 189
 190 HTTP clients that support the "smart" protocol (or both the
 191 "smart" and "dumb" protocols) MUST discover references by making
 192 a parameterized request for the info/refs file of the repository.
 193
 194 The request MUST contain exactly one query parameter,
 195 `service=$servicename`, where `$servicename` MUST be the service
 196 name the client wishes to contact to complete the operation.
 197 The request MUST NOT contain additional query parameters.
 198
 199    C: GET $GIT_URL/info/refs?service=git-upload-pack HTTP/1.0
 200
 201 dumb server reply:
 202
 203    S: 200 OK
 204    S:
 205    S: 95dcfa3633004da0049d3d0fa03f80589cbcaf31  refs/heads/maint
 206    S: d049f6c27a2244e12041955e262a404c7faba355  refs/heads/master
 207    S: 2cb58b79488a98d2721cea644875a8dd0026b115  refs/tags/v1.0
 208    S: a3c2e2402b99163d1d59756e5f207ae21cccba4c  refs/tags/v1.0^{}
 209
 210 smart server reply:
 211
 212    S: 200 OK
 213    S: Content-Type: application/x-git-upload-pack-advertisement
 214    S: Cache-Control: no-cache
 215    S:
 216    S: 001e# service=git-upload-pack\n
 217    S: 0000
 218    S: 004895dcfa3633004da0049d3d0fa03f80589cbcaf31 refs/heads/maint\0multi_ack\n
 219    S: 003fd049f6c27a2244e12041955e262a404c7faba355 refs/heads/master\n
 220    S: 003c2cb58b79488a98d2721cea644875a8dd0026b115 refs/tags/v1.0\n
 221    S: 003fa3c2e2402b99163d1d59756e5f207ae21cccba4c refs/tags/v1.0^{}\n
 222    S: 0000
 223
 224 The client may send Extra Parameters (see
 225 Documentation/technical/pack-protocol.txt) as a colon-separated string
 226 in the Git-Protocol HTTP header.
 227
 228 Dumb Server Response
 229 ^^^^^^^^^^^^^^^^^^^^
 230 Dumb servers MUST respond with the dumb server reply format.
 231
 232 See the prior section under dumb clients for a more detailed
 233 description of the dumb server response.
 234
 235 Smart Server Response
 236 ^^^^^^^^^^^^^^^^^^^^^
 237 If the server does not recognize the requested service name, or the
 238 requested service name has been disabled by the server administrator,
 239 the server MUST respond with the `403 Forbidden` HTTP status code.
 240
 241 Otherwise, smart servers MUST respond with the smart server reply
 242 format for the requested service name.
 243
 244 Cache-Control headers SHOULD be used to disable caching of the
 245 returned entity.
 246
 247 The Content-Type MUST be `application/x-$servicename-advertisement`.
 248 Clients SHOULD fall back to the dumb protocol if another content
 249 type is returned.  When falling back to the dumb protocol clients
 250 SHOULD NOT make an additional request to `$GIT_URL/info/refs`, but
 251 instead SHOULD use the response already in hand.  Clients MUST NOT
 252 continue if they do not support the dumb protocol.
 253
 254 Clients MUST validate the status code is either `200 OK` or
 255 `304 Not Modified`.
 256
 257 Clients MUST validate the first five bytes of the response entity
 258 matches the regex `^[0-9a-f]{4}#`.  If this test fails, clients
 259 MUST NOT continue.
 260
 261 Clients MUST parse the entire response as a sequence of pkt-line
 262 records.
 263
 264 Clients MUST verify the first pkt-line is `# service=$servicename`.
 265 Servers MUST set $servicename to be the request parameter value.
 266 Servers SHOULD include an LF at the end of this line.
 267 Clients MUST ignore an LF at the end of the line.
 268
 269 Servers MUST terminate the response with the magic `0000` end
 270 pkt-line marker.
 271
 272 The returned response is a pkt-line stream describing each ref and
 273 its known value.  The stream SHOULD be sorted by name according to
 274 the C locale ordering.  The stream SHOULD include the default ref
 275 named `HEAD` as the first ref.  The stream MUST include capability
 276 declarations behind a NUL on the first ref.
 277
 278 The returned response contains "version 1" if "version=1" was sent as an
 279 Extra Parameter.
 280
 281   smart_reply     =  PKT-LINE("# service=$servicename" LF)
 282                      "0000"
 283                      *1("version 1")
 284                      ref_list
 285                      "0000"
 286   ref_list        =  empty_list / non_empty_list
 287
 288   empty_list      =  PKT-LINE(zero-id SP "capabilities^{}" NUL cap-list LF)
 289
 290   non_empty_list  =  PKT-LINE(obj-id SP name NUL cap_list LF)
 291                      *ref_record
 292
 293   cap-list        =  capability *(SP capability)
 294   capability      =  1*(LC_ALPHA / DIGIT / "-" / "_")
 295   LC_ALPHA        =  %x61-7A
 296
 297   ref_record      =  any_ref / peeled_ref
 298   any_ref         =  PKT-LINE(obj-id SP name LF)
 299   peeled_ref      =  PKT-LINE(obj-id SP name LF)
 300                      PKT-LINE(obj-id SP name "^{}" LF
 301
 302
 303 Smart Service git-upload-pack
 304 ------------------------------
 305 This service reads from the repository pointed to by `$GIT_URL`.
 306
 307 Clients MUST first perform ref discovery with
 308 `$GIT_URL/info/refs?service=git-upload-pack`.
 309
 310    C: POST $GIT_URL/git-upload-pack HTTP/1.0
 311    C: Content-Type: application/x-git-upload-pack-request
 312    C:
 313    C: 0032want 0a53e9ddeaddad63ad106860237bbf53411d11a7\n
 314    C: 0032have 441b40d833fdfa93eb2908e52742248faf0ee993\n
 315    C: 0000
 316
 317    S: 200 OK
 318    S: Content-Type: application/x-git-upload-pack-result
 319    S: Cache-Control: no-cache
 320    S:
 321    S: ....ACK %s, continue
 322    S: ....NAK
 323
 324 Clients MUST NOT reuse or revalidate a cached response.
 325 Servers MUST include sufficient Cache-Control headers
 326 to prevent caching of the response.
 327
 328 Servers SHOULD support all capabilities defined here.
 329
 330 Clients MUST send at least one "want" command in the request body.
 331 Clients MUST NOT reference an id in a "want" command which did not
 332 appear in the response obtained through ref discovery unless the
 333 server advertises capability `allow-tip-sha1-in-want` or
 334 `allow-reachable-sha1-in-want`.
 335
 336   compute_request   =  want_list
 337                        have_list
 338                        request_end
 339   request_end       =  "0000" / "done"
 340
 341   want_list         =  PKT-LINE(want SP cap_list LF)
 342                        *(want_pkt)
 343   want_pkt          =  PKT-LINE(want LF)
 344   want              =  "want" SP id
 345   cap_list          =  capability *(SP capability)
 346
 347   have_list         =  *PKT-LINE("have" SP id LF)
 348
 349 TODO: Document this further.
 350
 351 The Negotiation Algorithm
 352 ~~~~~~~~~~~~~~~~~~~~~~~~~
 353 The computation to select the minimal pack proceeds as follows
 354 (C = client, S = server):
 355
 356 'init step:'
 357
 358 C: Use ref discovery to obtain the advertised refs.
 359
 360 C: Place any object seen into set `advertised`.
 361
 362 C: Build an empty set, `common`, to hold the objects that are later
 363    determined to be on both ends.
 364
 365 C: Build a set, `want`, of the objects from `advertised` the client
 366    wants to fetch, based on what it saw during ref discovery.
 367
 368 C: Start a queue, `c_pending`, ordered by commit time (popping newest
 369    first).  Add all client refs.  When a commit is popped from
 370    the queue its parents SHOULD be automatically inserted back.
 371    Commits MUST only enter the queue once.
 372
 373 'one compute step:'
 374
 375 C: Send one `$GIT_URL/git-upload-pack` request:
 376
 377    C: 0032want <want #1>...............................
 378    C: 0032want <want #2>...............................
 379    ....
 380    C: 0032have <common #1>.............................
 381    C: 0032have <common #2>.............................
 382    ....
 383    C: 0032have <have #1>...............................
 384    C: 0032have <have #2>...............................
 385    ....
 386    C: 0000
 387
 388 The stream is organized into "commands", with each command
 389 appearing by itself in a pkt-line.  Within a command line,
 390 the text leading up to the first space is the command name,
 391 and the remainder of the line to the first LF is the value.
 392 Command lines are terminated with an LF as the last byte of
 393 the pkt-line value.
 394
 395 Commands MUST appear in the following order, if they appear
 396 at all in the request stream:
 397
 398 * "want"
 399 * "have"
 400
 401 The stream is terminated by a pkt-line flush (`0000`).
 402
 403 A single "want" or "have" command MUST have one hex formatted
 404 object name as its value.  Multiple object names MUST be sent by sending
 405 multiple commands. Object names MUST be given using the object format
 406 negotiated through the `object-format` capability (default SHA-1).
 407
 408 The `have` list is created by popping the first 32 commits
 409 from `c_pending`.  Less can be supplied if `c_pending` empties.
 410
 411 If the client has sent 256 "have" commits and has not yet
 412 received one of those back from `s_common`, or the client has
 413 emptied `c_pending` it SHOULD include a "done" command to let
 414 the server know it won't proceed:
 415
 416    C: 0009done
 417
 418 S: Parse the git-upload-pack request:
 419
 420 Verify all objects in `want` are directly reachable from refs.
 421
 422 The server MAY walk backwards through history or through
 423 the reflog to permit slightly stale requests.
 424
 425 If no "want" objects are received, send an error:
 426 TODO: Define error if no "want" lines are requested.
 427
 428 If any "want" object is not reachable, send an error:
 429 TODO: Define error if an invalid "want" is requested.
 430
 431 Create an empty list, `s_common`.
 432
 433 If "have" was sent:
 434
 435 Loop through the objects in the order supplied by the client.
 436
 437 For each object, if the server has the object reachable from
 438 a ref, add it to `s_common`.  If a commit is added to `s_common`,
 439 do not add any ancestors, even if they also appear in `have`.
 440
 441 S: Send the git-upload-pack response:
 442
 443 If the server has found a closed set of objects to pack or the
 444 request ends with "done", it replies with the pack.
 445 TODO: Document the pack based response
 446
 447    S: PACK...
 448
 449 The returned stream is the side-band-64k protocol supported
 450 by the git-upload-pack service, and the pack is embedded into
 451 stream 1.  Progress messages from the server side MAY appear
 452 in stream 2.
 453
 454 Here a "closed set of objects" is defined to have at least
 455 one path from every "want" to at least one "common" object.
 456
 457 If the server needs more information, it replies with a
 458 status continue response:
 459 TODO: Document the non-pack response
 460
 461 C: Parse the upload-pack response:
 462    TODO: Document parsing response
 463
 464 'Do another compute step.'
 465
 466
 467 Smart Service git-receive-pack
 468 ------------------------------
 469 This service reads from the repository pointed to by `$GIT_URL`.
 470
 471 Clients MUST first perform ref discovery with
 472 `$GIT_URL/info/refs?service=git-receive-pack`.
 473
 474    C: POST $GIT_URL/git-receive-pack HTTP/1.0
 475    C: Content-Type: application/x-git-receive-pack-request
 476    C:
 477    C: ....0a53e9ddeaddad63ad106860237bbf53411d11a7 441b40d833fdfa93eb2908e52742248faf0ee993 refs/heads/maint\0 report-status
 478    C: 0000
 479    C: PACK....
 480
 481    S: 200 OK
 482    S: Content-Type: application/x-git-receive-pack-result
 483    S: Cache-Control: no-cache
 484    S:
 485    S: ....
 486
 487 Clients MUST NOT reuse or revalidate a cached response.
 488 Servers MUST include sufficient Cache-Control headers
 489 to prevent caching of the response.
 490
 491 Servers SHOULD support all capabilities defined here.
 492
 493 Clients MUST send at least one command in the request body.
 494 Within the command portion of the request body clients SHOULD send
 495 the id obtained through ref discovery as old_id.
 496
 497   update_request  =  command_list
 498                      "PACK" <binary data>
 499
 500   command_list    =  PKT-LINE(command NUL cap_list LF)
 501                      *(command_pkt)
 502   command_pkt     =  PKT-LINE(command LF)
 503   cap_list        =  *(SP capability) SP
 504
 505   command         =  create / delete / update
 506   create          =  zero-id SP new_id SP name
 507   delete          =  old_id SP zero-id SP name
 508   update          =  old_id SP new_id SP name
 509
 510 TODO: Document this further.
 511
 512
 513 References
 514 ----------
 515
 516 http://www.ietf.org/rfc/rfc1738.txt[RFC 1738: Uniform Resource Locators (URL)]
 517 http://www.ietf.org/rfc/rfc2616.txt[RFC 2616: Hypertext Transfer Protocol -- HTTP/1.1]
 518 link:technical/pack-protocol.html
 519 link:technical/protocol-capabilities.html