On Thu, Feb 1, 2018 at 10:48 AM, Jeff Hostetler [off-list ref] wrote:

On 1/2/2018 7:18 PM, Brandon Williams wrote:

quoted

Introduce git-serve, the base server for protocol version 2.

Protocol version 2 is intended to be a replacement for Git's current
wire protocol.  The intention is that it will be a simpler, less
wasteful protocol which can evolve over time.

Protocol version 2 improves upon version 1 by eliminating the initial
ref advertisement.  In its place a server will export a list of
capabilities and commands which it supports in a capability
advertisement.  A client can then request that a particular command be
executed by providing a number of capabilities and command specific
parameters.  At the completion of a command, a client can request that
another command be executed or can terminate the connection by sending a
flush packet.

Signed-off-by: Brandon Williams <redacted>
---
  .gitignore                              |   1 +
  Documentation/technical/protocol-v2.txt |  91 ++++++++++++
  Makefile                                |   2 +
  builtin.h                               |   1 +
  builtin/serve.c                         |  30 ++++
  git.c                                   |   1 +
  serve.c                                 | 239
++++++++++++++++++++++++++++++++
  serve.h                                 |  15 ++
  8 files changed, 380 insertions(+)
  create mode 100644 Documentation/technical/protocol-v2.txt
  create mode 100644 builtin/serve.c
  create mode 100644 serve.c
  create mode 100644 serve.h

diff --git a/.gitignore b/.gitignore
index 833ef3b0b..2d0450c26 100644
--- a/.gitignore
+++ b/.gitignore

@@ -140,6 +140,7 @@
  /git-rm
  /git-send-email
  /git-send-pack
+/git-serve
  /git-sh-i18n
  /git-sh-i18n--envsubst
  /git-sh-setup

diff --git a/Documentation/technical/protocol-v2.txt

b/Documentation/technical/protocol-v2.txt
new file mode 100644
index 000000000..b87ba3816

--- /dev/null
+++ b/Documentation/technical/protocol-v2.txt

@@ -0,0 +1,91 @@
+ Git Wire Protocol, Version 2
+==============================
+
+This document presents a specification for a version 2 of Git's wire
+protocol.  Protocol v2 will improve upon v1 in the following ways:
+
+  * Instead of multiple service names, multiple commands will be
+    supported by a single service.
+  * Easily extendable as capabilities are moved into their own section
+    of the protocol, no longer being hidden behind a NUL byte and
+    limited by the size of a pkt-line (as there will be a single
+    capability per pkt-line).
+  * Separate out other information hidden behind NUL bytes (e.g. agent
+    string as a capability and symrefs can be requested using 'ls-refs')
+  * Reference advertisement will be omitted unless explicitly requested
+  * ls-refs command to explicitly request some refs
+
+ Detailed Design
+=================
+
+A client can request to speak protocol v2 by sending `version=2` in the
+side-channel `GIT_PROTOCOL` in the initial request to the server.
+
+In protocol v2 communication is command oriented.  When first contacting

a
+server a list of capabilities will advertised.  Some of these
capabilities
+will be commands which a client can request be executed.  Once a command
+has completed, a client can reuse the connection and request that other
+commands be executed.
+
+ Special Packets
+-----------------
+
+In protocol v2 these special packets will have the following semantics:
+
+  * '0000' Flush Packet (flush-pkt) - indicates the end of a message
+  * '0001' Delimiter Packet (delim-pkt) - separates sections of a message

Previously, a 0001 pkt-line meant that there was 1 byte of data
following, right?

No, the length was including the length field, so 0005 would indicate that
there is one byte following, (+4 bytes of "0005" included)

Does this change that and/or prevent 1 byte
packets?  (Not sure if it is likely, but the odd-tail of a packfile
might get sent in a 0001 line, right?)  Or is it that 0001 is only
special during the V2 negotiation stuff, but not during the packfile
transmission?

0001 is invalid in the current protocol v0.

(I'm not against having this delimiter -- I think it is useful, but
just curious if will cause problems elsewhere.)

Should we also consider increasing the pkt-line limit to 5 hex-digits
while we're at it ?   That would let us have 1MB buffers if that would
help with large packfiles.

AFAICT there is a static allocation of one pkt-line (of maximum size),
such that the code can read in a full packet and then process it.
If we'd increase the packet size we'd need the static buffer to be 1MB,
which sounds good for my developer machine. But I suspect it may be
too much for people using git on embedded devices?

pack files larger than 64k are put into multiple pkt-lines, which is
not a big deal, as the overhead of 4bytes per 64k is negligible.
(also there is progress information in the side channel, which
would come in as a special packet in between real packets,
such that every 64k transmitted you can update your progress
meter; Not sure I feel strongly on fewer progress updates)

 Granted, we're throttled by the network,
so it might not matter.  Would it be interesting to have a 5 digit
prefix with parts of the high bits of first digit being flags ?
Or is this too radical of a change?

What would the flags be for?

As an alternative we could put the channel number in one byte,
such that we can have a side channel not just while streaming the
pack but all the time. (Again, not sure if that buys a lot for us)