summaryrefslogtreecommitdiffstats
path: root/doc/TODO
blob: d58bc3aba69dd34fc4491834c7616111cc9eeb79 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
Important for 1.0
===

 - Audit @todo and FIXME for security issues

 - The convenience functions in dbus-bus.h should perhaps have
   the signatures that they would have if they were autogenerated
   stubs. e.g. the acquire service function. We should also evaluate 
   which of these functions to include, in light of the fact that 
   GLib/Qt native stubs will probably also exist.

 - the "break loader" and valid/invalid message tests are all disabled;
   they need to be fixed and re-enabled with the new message args stuff.
   I think I want to drop the .message files thing and just have code
   that generates messages, more like the tests for
   dbus-marshal-recursive.c (this is mostly done now, just needs some
   cleanup)

 - need to define bus behavior if you send a message to 
   yourself; is it an error, or allowed? If allowed, 
   we need to have a test for it in the test suite.

 - validate dict entry number of fields

 - just before 1.0, try a HAVE_INT64=0 build and be sure it runs

 - in dbus-keyring.c, enforce that the keyring dir is not 
   world readable/writable

 - Ping isn't handled

 - fix introspection format to handle all signatures

 - make the mutex/condvar functions private

 - dbus-pending-call.c has some API and thread safety issues to review

 - make dbus_connection_open() return a shared connection from a pool.
   Add dbus_connection_open_private() that works like the current one.
   To do this, each DBusServer could have a 128-bit GUID. This GUID
   would be in the address from dbus_server_get_address(). On
   connection to a server, the GUID would be provided as the first
   thing in the auth protocol, and verified vs. the expected GUID if a
   GUID was in the address used to connect. A hash from GUID to
   connection would be kept, so attempts to connect to a GUID already
   in the hash would return a shared existing connection.

   The purpose of all this is to allow a dbus_g_proxy_to_string() that
   would convert the proxy to an "IOR" and dbus_g_proxy_from_string()
   that would decode; using these, dbus-glib users could avoid
   DBusConnection entirely. Of course the same applies to other kinds
   of binding.

Important for 1.0 GLib Bindings
===

 - finish dbus-glib-tool support for adding introspection 
   data to GObject and autoexporting GObject using same

 - Need to make sure that dbus-glib.h never returns any 
   dbus_malloc() memory, only g_malloc(). 
   dbus_g_proxy_end_call() is the major offender. 

 - DBusGProxy doesn't emit "destroy" when it should

Might as Well for 1.0
===

 - add dbus_message_has_path(), maybe has_member/interface

 - connection_open/connection_disconnect lacks symmetry, open/close
   or connect/disconnect

 - protocol version in each message is pretty silly

Can Be Post 1.0
===

 - DBusWatchList/TimeoutList duplicate a lot of code, as do
   protected_change_watch/protected_change_timeout in dbus-connection.c
   and dbus-server.c. This could all be mopped up, cut-and-paste 
   fixed, code size reduced.

 - change .service files to allow Names=list in addition to Name=string

 - The message bus internal code still says "service" for 
   "name", "base service" for "unique name", "activate" for 
   "start"; would be nice to clean up.

 - Property list feature on message bus (list of properties associated 
   with a connection). May also include message matching rules 
   that involve the properties of the source or destination
   connection.

 - Disconnecting the remote end on invalid UTF-8 is probably not a good 
   idea. The definition of "valid" is slightly fuzzy. I think it might 
   be better to just silently "fix" the UTF-8, or perhaps return an error.

 - build and install the Doxygen manual in Makefile when --enable-docs

 - if you send the same message to multiple connections, the serial number 
   will only be right for one of them. Probably need to just write() the serial 
   number, rather than putting it in the DBusMessage, or something.

 - perhaps the bus driver should have properties that reflect attributes
   of the session, such as hostname, architecture, operating system, 
   etc. Could be useful for code that wants to special-case behavior 
   for a particular host or class of hosts, for example.

 - currently the security policy stuff for messages to/from 
   the bus driver is kind of strange; basically it's hardcoded that 
   you can always talk to the driver, but the default config file 
   has rules for it anyway, or something. it's conceptually 
   screwy at the moment.

 - when making a method call, if the call serial were globally unique,
   we could forward the call serial along with any method calls made
   as a result of the first method call, and allow reentrancy that was
   strictly part of the call stack of said method call. But I don't
   really see how to do this without making the user pass around the
   call serial to all method calls all the time, or disallowing 
   async calls.

   If done post 1.0 will probably be an optional/ugly-API type 
   of thing.

 - I don't want to introduce DBusObject, but refcounting and object
   data could still be factored out into an internal "base class" 
   perhaps.

 - document the auth protocol as a set of states and transitions, and
   then reimplement it in those terms

 - recursive dispatch, see dbus_connection_dispatch()

 - do we need per-display activation; if so I'd like to do this by setting a
   "display ID" property on screen 0, with a GUID, and keying activation by 
   said GUID. Otherwise you get all kinds of unrobust
   string/hostname-based mess. per-screen is then done by appending screen number
   to the display. If displays have a deterministic ID like this, you can 
   do per-display by simply including GUID in the service name.

 - optimization and profiling!