chrome://plugins Mojo-ification part 1/2, adding interface.

chrome/browser/ui/webui/plugins/plugins.mojom

Index: chrome/browser/ui/webui/plugins/plugins.mojom
diff --git a/chrome/browser/ui/webui/plugins/plugins.mojom b/chrome/browser/ui/webui/plugins/plugins.mojom
new file mode 100644
index 0000000000000000000000000000000000000000..17c435a6cb40c0eb269339032383737acb1904ab
--- /dev/null
+++ b/chrome/browser/ui/webui/plugins/plugins.mojom
@@ -0,0 +1,45 @@
+// Copyright 2015 The Chromium Authors. All rights reserved.

[sky, 2016/01/22 00:58:34]

2016

[dpapad, 2016/01/22 02:00:06]

Done.

+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+struct MimeType {
+  string description;
+  array<string> file_extensions;
+  string mime_type;
+};
+
+struct PluginFile {
+  string description;
+  string enabled_mode;
+  array<MimeType> mime_types;
+  string name;
+  string path;
+  string type;
+  string version;
+};
+
+struct PluginData {
+  bool always_allowed;
+  string description;
+  bool critical;
+  string enabled_mode;
+  string id;
+  string name;
+  string update_url;
+  string version;
+  array<PluginFile> plugin_files;
+};
+
+interface PluginsHandlerMojo {
+  GetPluginsData() => (array<PluginData> plugins);

[sky, 2016/01/22 00:58:34]

Do you need both GetPluginsData() and OnPluginsUpdated()? I'm wondering if you
make SetClientPage call OnPluginsUpdated() then you shouldn't need the
GetPluginsData?

[dpapad, 2016/01/22 02:00:06]

The JS code is calling GetPluginsData() not only on startup but also within a
recurring 30sec timer, see follow up CL
https://codereview.chromium.org/1576183003 (previous code also did this), so I
don't think I can remove GetPluginsData(). BTW, I am re-using as much code as
possible both in C++ and JS (see follow up CL). 

But still, you raise a good point about this API design dilemma here that will
probably come up in other Mojo interfaces.

My thinking is as follows: One of the benefits of using Mojo is the explicit
association of a request (JS->C++) and a response (C++->JS) using Promises. The
suggested change would make the API resemble more the chrome.send() paradigm,
where a request is sent, and then a method is called as a callback, but the
association between request/response is hidden (documented in comments usually).
Making it explicit that one codepath deals with a direct response and one deals
with plugin info updated from a different chrome://plugins tab seems more
intuitive to me. WDYT?

+  GetShowDetails() => (bool show_details);
+  SaveShowDetailsToPrefs(bool details_mode);
+  SetPluginAlwaysAllowed(string plugin, bool allowed);
+  SetPluginEnabled(string plugin_path, bool enable);
+  SetPluginGroupEnabled(string group_name, bool enable);
+  SetClientPage(PluginsPageMojo page);
+};
+
+interface PluginsPageMojo {
+  OnPluginsUpdated(array<PluginData> plugins);
+};