From 3e94cafa43d464d73815e21810b97a4faf54136d Mon Sep 17 00:00:00 2001
From: drebs <drebs@leap.se>
Date: Sun, 18 Jun 2017 11:18:10 -0300
Subject: [pkg] unify client and server into a single python package

We have been discussing about this merge for a while.
Its main goal is to simplify things: code navigation, but also
packaging.

The rationale is that the code is more cohesive in this way, and there's
only one source package to install.

Dependencies that are only for the server or the client will not be
installed by default, and they are expected to be provided by the
environment. There are setuptools extras defined for the client and the
server.

Debianization is still expected to split the single source package into
3 binaries.

Another avantage is that the documentation can now install a single
package with a single step, and therefore include the docstrings into
the generated docs.

- Resolves: #8896
---
 src/leap/soledad/common/README.txt | 70 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 70 insertions(+)
 create mode 100644 src/leap/soledad/common/README.txt

(limited to 'src/leap/soledad/common/README.txt')

diff --git a/src/leap/soledad/common/README.txt b/src/leap/soledad/common/README.txt
new file mode 100644
index 00000000..0a252650
--- /dev/null
+++ b/src/leap/soledad/common/README.txt
@@ -0,0 +1,70 @@
+Soledad common package
+======================
+
+This package contains Soledad bits used by both server and client.
+
+Couch L2DB Backend
+------------------
+
+L2DB backends rely on some atomic operations that modify documents contents
+and metadata (conflicts, transaction ids and indexes). The only atomic
+operation in Couch is a document put, so every u1db atomic operation has to be
+mapped to a couch document put.
+
+The atomic operations in the U1DB SQLite reference backend implementation may
+be identified by the use of a context manager to access the underlying
+database. A listing of the methods involved in each atomic operation are
+depiced below. The top-level elements correpond to the atomic operations that
+have to be mapped, and items on deeper levels of the list have to be
+implemented in a way that all changes will be pushed with just one operation.
+
+    * _set_replica_uid
+    * put_doc:
+        * _get_doc
+        * _put_and_update_indexes
+            * insert/update the document
+            * insert into transaction log
+    * delete_doc
+        * _get_doc
+        * _put_and_update_indexes
+    * get_doc_conflicts
+        * _get_conflicts
+    * _set_replica_gen_and_trans_id
+        * _do_set_replica_gen_and_trans_id
+    * _put_doc_if_newer
+        * _get_doc
+        * _validate_source (**)
+            * _get_replica_gen_and_trans_id
+        * cases:
+            * is newer:
+                * _prune_conflicts (**)
+                    * _has_conflicts
+                    * _delete_conflicts
+                * _put_and_update_indexes
+            * same content as:
+                * _put_and_update_indexes
+            * conflicted:
+                * _force_doc_sync_conflict
+                    * _prune_conflicts
+                    * _add_conflict
+                    * _put_and_update_indexes
+        * _do_set_replica_gen_and_trans_id
+    * resolve_doc
+        * _get_doc
+        * cases:
+            * doc is superseded
+                * _put_and_update_indexes
+            * else
+                * _add_conflict
+        * _delete_conflicts
+    * delete_index
+    * create_index
+
+Notes:
+
+  * Currently, the couch backend does not implement indexing, so what is
+    depicted as `_put_and_update_indexes` above will be found as `_put_doc` in
+    the backend.
+
+  * Conflict updates are part of document put using couch update functions,
+    and as such are part of the same atomic operation as document put.
-- 
cgit v1.2.3