From 725119ae497e4f8d46cbc04c0d97bccfa84ad4f1 Mon Sep 17 00:00:00 2001
From: Gerrit User 1689 <1689@4a232e18-c5a9-48ee-94c0-e04e7cca6543>
Date: Thu, 17 Apr 2014 20:04:02 +0000
Subject: [PATCH] Update patch set 1

Patch Set 1: (1 inline comment)

Some thoughts (not even nits) in-line.

Patch-set: 1
Reviewer: Gerrit User 1689 <1689@4a232e18-c5a9-48ee-94c0-e04e7cca6543>
Label: Code-Review=0
Label: Workflow=0
---
 81a3267ea6d7dcf41a6eeb7a062592a3735bfb9f | 21 +++++++++++++++++++++
 1 file changed, 21 insertions(+)
 create mode 100644 81a3267ea6d7dcf41a6eeb7a062592a3735bfb9f

diff --git a/81a3267ea6d7dcf41a6eeb7a062592a3735bfb9f b/81a3267ea6d7dcf41a6eeb7a062592a3735bfb9f
new file mode 100644
index 000000000..e8522789b
--- /dev/null
+++ b/81a3267ea6d7dcf41a6eeb7a062592a3735bfb9f
@@ -0,0 +1,21 @@
+{
+  "comments": [
+    {
+      "key": {
+        "uuid": "AAAAXH//vLk\u003d",
+        "filename": "specs/juno/l3-svcs-vendor-validation.rst",
+        "patchSetId": 1
+      },
+      "lineNbr": 54,
+      "author": {
+        "id": 1689
+      },
+      "writtenOn": "2014-04-17T20:04:02Z",
+      "side": 1,
+      "message": "Just some thoughts to consider, not necessarily anything that has to be changed in this BP:\n\nThese phases seem to me to be somewhat similar to the ML2 core plugin\u0027s pattern of invoking precommit() and postcommit() operations on the mechanism drivers for each create, update, and delete operation. But ML2 uses only two phases. It\u0027s precommit() is called within the DB transaction, and encompasses both the validate and persist steps being proposed here. Exceptions thrown from precommit() methods by drivers result in the entire transaction being rolled back and an exception returned to the client. ML2\u0027s postcommit() happens after the transaction has been committed, and seems to correspond to your apply phase.\n\nYou may want to reconsider whether the validate and persist phases need to be separate methods. One thing that unifying these allows is for \"validation\" to include reserving any needed resources from pools at the DB level.\n\nIf you do decide that two phases at the driver level is sufficent, and the semantics of the driver methods you end up with match those in ML2, I\u0027d recommend trying to keep the method naming and other details consistent with ML2 as well.",
+      "revId": "81a3267ea6d7dcf41a6eeb7a062592a3735bfb9f",
+      "serverId": "4a232e18-c5a9-48ee-94c0-e04e7cca6543",
+      "unresolved": false
+    }
+  ]
+}
\ No newline at end of file