File: INTERNALS

package info (click to toggle)
microdc2 0.15.6-7
  • links: PTS, VCS
  • area: main
  • in suites: sid, trixie
  • size: 4,620 kB
  • sloc: ansic: 31,186; sh: 4,024; makefile: 142; sed: 16
file content (186 lines) | stat: -rw-r--r-- 7,021 bytes parent folder | download | duplicates (5)
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
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
This document contains various technical information geared towards
developers of microdc.

===

User connection identification improvements:
  There are four possible cases to consider:

  We initiating       | They initiating
  =================================================
  !+RCTM (we passive) | %+RCTM => we initiating CTM
  %+CTM  (we active)  |  -CTM

  (R)CTM = (Rev)ConnectToMe
  -  = We cannot antipicate
  !+ = We can anticipate by pending_userinfo and we_connected=true
  %+ = We can anticipate by pending_userinfo and we_connected=false

  Current implementation has pending_userinfo. To be more accurate,
  we should split this map into pending_userinfo_{in,out}.
  validate_nick need to be modified to check based on uc->we_connected.

RevConnectToMe vs ConnectToMe considerations:
  1) Security consideration: We don't want to connect to other users.
     It's the one that connects that starts sending data. The remote user
     could make us connect and send data ($MyNick+$Lock) to anyone on the
     net, which is bad.
  2) Some users may think they are active, but they are not.

  The following is only relevant for active microdc users:
  These two considerations gives that it is better to be connected to than
  connect to, i.e. it is better to send ConnectToMe's and receive
  RevConnectToMe's rather than to send RevConnectToMe's and receive
  ConnectToMe's.

  So here's one an improved scheme for response to remote connection
  initiations.

    1. Receive ConnectToMe from remote.
    2. Ignore ConnectToMe from remote, and send ConnectToMe to remote.
    3.1. If remote is not running microdc, remote will now connect to us.
    3.2. If remote is running microdc, remote will send ConnectToMe to us again.
    4.2. We connect to the remote.

  Problems:
    * Hubs may block repeated ConnectToMe's.
    * Clients' behavior on receiving ConnectToMe right after sending
      ConnectToMe could be unexpected.
    * This scheme is unfair to other clients not implementing this scheme.

User active/passive state transition
  At the moment, there are four values for DCActiveState (DCUserInfo.active_state).
  Initial state is UNKNOWN.

  Note. KNOWN_ACTIVE only means that the user states that he is active.
        It means that the user's client has active enabled.
        But the client may be misconfigured, or the user's firewall may be.
  Note. RECEIVED_PASSIVE doesn't mean that the user states that he is passive.
        It only means that the user sent us $RevConnectToMe. It is unclear
	whether the client is forbidden to send $RevConnectToMe in
	active mode.

  UNKNOWN
    set on SENT_ACTIVE       in main.c:DC_MSG_VALIDATE_NICK:ok                       # make way for future ConnectToMe
  KNOWN_ACTIVE
    set on SENT_PASSIVE      in main.c:DC_MSG_VALIDATE_NICK:ok                       # logically!
  RECEIVED_PASSIVE
    set on *                 in hub.c:hub_handle_command:$RevConnectToMe             # naturally!
    tested for               in hub.c:hub_handle_command:$RevConnectToMe:!is_active  # do we need to send back $RevConnectToMe?
    tested for               in hub.c:hub_connect_user:!is_active                    # user also passive
  SENT_PASSIVE
    tested for               in hub.c:hub_handle_command:$RevConnectToMe:!is_active  # cannot communicate?
    tested for               in hub.c:hub_connect_user:!is_active                    # waiting - RevConnectToMe already sent
    set on !RECEIVED_PASSIVE in hub.c:hub_connect_user:!is_active
  SENT_ACTIVE
    tested for               in hub.c:hub_connect_user:is_active                     # waiting - ConnectToMe already sent
    set on *                 in hub.c:hub_connect_user:is_active

  When passive:
    UNKNOWN => SENT_PASSIVE => RECEIVED_PASSIVE [DC++ clients]
    UNKNOWN => SENT_PASSIVE [non-DC++ clients]
    UNKNOWN => SENT_PASSIVE => KNOWN_ACTIVE => SENT_PASSIVE => ..
    UNKNOWN => RECEIVED_PASSIVE
  When active:
    UNKNOWN => RECEIVED_PASSIVE => SENT_ACTIVE => UNKNOWN => ..
    UNKNOWN => SENT_ACTIVE => UNKNOWN => ..

  When user sends new $MyINFO
    RECEIVED_PASSIVE => UNKNOWN
    KNOWN_ACTIVE => UNKNOWN
    SENT_PASSIVE unchanged - technically we could try resending this. XXX?
    SENT_ACTIVE unchanged - will always work

  Problem: Loss of KNOWN_ACTIVE state
    We lose KNOWN_ACTIVE knowledge if
    1) we are passive (!is_active)
    2) we have connected to remote user, and state has become KNOWN_ACTIVE
    3) we try to connect to remote user again by sending $RevConnectToMe
    State becomes SENT_PASSIVE.

  Problem: Loss of RECEIVED_PASSIVE state
    Assume we are passive. State for some user is RECEIVED_PASSIVE.
    We then become active (setting changed). RECEIVED_PASSIVE information is
    when we try to connect to user next time.
  Since passive DC++ clients only sends RevConnectToMe back once in reply
  to $RevConnectToMe, it means we might get stuck in SENT_PASSIVE state.
  This is not a problem though.

  Relevant states in microdc active mode:  all but SENT_ACTIVE
  Relevant states in microdc passive mode: all but SEND_PASSIVE
  When changing from active to passive, we must only consider SENT_ACTIVE state.
    OK : SENT_ACTIVE tested for in hub.c:hub_connect_user:is_active -- is_active is false!
  When changing from passive to active, we must only consider SENT_PASSIVE state.
    OK : SENT_PASSIVE tested for in hub.c:hub_connect_user:!is_active -- is_active is true! 
    OK : SENT_PASSIVE tested for in hub.c:hub_handle_command:$RevConnectToMe:!is_active -- cannot communicate?
    OK : KNOWN_ACTIVE set on SENT_PASSIVE in main.c:DC_MSG_VALIDATE_NICK:ok -- this is still true.

===

Coding notes:
  Error management:
    Assume ((x)v)asprintf only fails due to memory exhaustion
      (assuming format appears to be OK) or if it fails, just abort()
    Handle memory exhaustion errors by dying immediately
    Handle every other error that can occur!

===

User commands:

  $MyNick		s     		MYNICK
  $Direction		s i 		DIRECTION
  $Error		s...  		*
  $FileLength		i     		FILE_LENGTH
  $GetListLen		-	    	*
  $ListLen		i     		*   	    	sent as a reply to $GetListLen
  $Get			s$i 		GET
  $GetTestZBlock	?	    	*
  $GetZBlock		?	    	*
  $UGetZBlock		?	    	*
  $UGetBlock		?	    	*
  $Key			s...  		KEY
  $Lock			s...  		LOCK
  $Send			-	    	SEND
  $Sending		[i]   		FILE_LENGTH
  $MaxedOut		-	    	FILE_LENGTH
  $Supports		s...  		*

Hub commands:

  $ValidateNick		s
  $Key			s...
  $Version		s...
  $GetNickList		-
  $MyPass		s...
  $GetINFO		s
  <			s> s...
  $ConnectToMe		s s:i
  $To:			s From: s $s
  $Supports		s...
  $RevConnectToMe	s s
  $OpForceMove		$Who:s$Where:s$Msg:s
  $Search		?
  $MyINFO		?
  $Quit			?
  $SR			?
  $HubName		s...
  $Supports		?
  $UserCommand		?
  $Lock			s Pk=s
  $Hello		s
  $ForceMove		?
  $HubIsFull		?
  $ValidateDenide	-
  $UserIP		?
  $NickList		s{$$s}
  $OpList		[s{$$s}]
  $GetPass		-
  $BadPass		-
  $LogedIn		-
  $Kick			?
  $Capabilities		[s{$s}]			dchub:YES dc++:NO

===

curses client