| 1 | Revision history for AnyEvent::MP |
1 | Revision history for AnyEvent::MP |
| 2 | |
2 | |
| 3 | TODO: global protocol fix? |
|
|
| 4 | TODO: testsuite |
3 | TODO: testsuite |
| 5 | TODO: make more parameters configurable: framing, tls, auth, timeouts |
|
|
| 6 | TODO: intro: maybe simple job pool example? |
4 | TODO: intro: maybe simple job pool example? |
| 7 | TODO: intro: mention watchdog... |
5 | TODO: intro: mention watchdog... |
| 8 | TODO: aemp manpage more examples |
6 | TODO: maybe disbale warnings by default? |
| 9 | TODO: group monitoring |
7 | TODO: listener-scopes (10.0.0.1:4040@vpn) and connect-scopes ("vpn,public") |
| 10 | |
8 | |
| 11 | TODO: cal, cal_func |
9 | faq: can't se anything |
|
|
10 | faq: all is asynch |
|
|
11 | faq: how to interface to non-perl nodes? |
|
|
12 | |
|
|
13 | TODO: update docs |
|
|
14 | TODO: switch to AnyEvent::Log |
|
|
15 | TODO: check gproto, nproto, on connect |
|
|
16 | TODO: maybe switch to md5+sha1+sha256 or so... |
|
|
17 | TODO: up_nodes for aemp-trace replace by rpc call? |
|
|
18 | TODO: listener-less nodes - not supported anymore |
|
|
19 | TODO: eg/* |
|
|
20 | TODO: secure node request |
|
|
21 | TODO: timeout |
|
|
22 | TODO: porting-guide |
|
|
23 | TODO: global undocumented compat functions? |
|
|
24 | TODO: #TODO##d# markers |
|
|
25 | TODO: make sRCNODE efficient and the nodename (bind send/recv late) |
|
|
26 | 2 |
|
|
27 | - INCOMPATIBLE CHANGE: the node protocol used by this version is not |
|
|
28 | compatible to versions <2 - you have to upgrade all nodes. |
|
|
29 | - INCOMPATIBLE CHANGE: many APIs have changed, see the section |
|
|
30 | "PORTING FROM VERSION 1" in the AnyEvent::MP manpage for details. |
|
|
31 | - AnyEvent::MP::Global is now optional - only seed nodes autoload it |
|
|
32 | and create a fully-meshed network with each other, normal nodes only |
|
|
33 | have to connect to seed nodes, other connections are on-demand. |
|
|
34 | - new "norc" and "force" options for MP::configure to ignore |
|
|
35 | or override the rc file. |
|
|
36 | - move seed code to Kernel. |
|
|
37 | - fixed small typo in MP::Transport. |
|
|
38 | - fix (unused) provider-kv (version sometimes missing). |
|
|
39 | - reduce the default export list of MP::Kernel. |
|
|
40 | - reduced various random delays to zero, as they |
|
|
41 | are hopefully no longer needed with the new global |
|
|
42 | protocol. |
|
|
43 | - implement a secure mode that can suppress remote code execution |
|
|
44 | requests on a per-node/per-request basis. |
|
|
45 | - do not use GMP for generating alphanumeric strings even when |
|
|
46 | available, use a more tricky algorithm instead. |
|
|
47 | - anonymous node names and remote port names have been shortened |
|
|
48 | considerably. |
|
|
49 | - aemp shell now supports "package" selection and offers an |
|
|
50 | $ECHO port you cna send messages to. |
|
|
51 | - rely on perl rand() instead of using /dev/urandom when available, as |
|
|
52 | hopefully it is random enough. |
|
|
53 | - aemp shell no longer leaks one port per command. |
|
|
54 | - nodenames can contain %-escapes such as %n or %u. |
|
|
55 | - aemp now uses aemp/%n/%u as nodename. |
|
|
56 | - known_nodes is gone, it has little value, use all_nodes as |
|
|
57 | replacement. |
|
|
58 | - lots of small change and improvements. |
|
|
59 | - TRACE now shows the portid on SND. |
|
|
60 | |
|
|
61 | 1.30 Thu Jun 30 11:30:39 CEST 2011 |
|
|
62 | - connection errors at the right time would kill node |
|
|
63 | connections in a bad way (patch by Malcolm Studd). |
|
|
64 | |
|
|
65 | 1.29 Fri May 7 20:13:39 CEST 2010 |
|
|
66 | - codename "Sadrak". |
|
|
67 | - fix error in callback at AnyEvent/MP/Global.pm line 339 |
|
|
68 | (found by Sadrak). |
|
|
69 | - listener-less nodes had trouble sending keepalive |
|
|
70 | messages on write timeouts (found by Sadrak). |
|
|
71 | - the monitor guard could cause a memleak due |
|
|
72 | to autovivification if a mon was cleared after |
|
|
73 | the port was gone (analysed by Sadrak). |
|
|
74 | - do not overwrite the config file if we couldn't read it |
|
|
75 | for some reason (lesson demonstrated by Sadrak). |
|
|
76 | |
|
|
77 | 1.28 Thu Apr 1 21:23:54 CEST 2010 |
|
|
78 | - accepted connections didn't correctly set up the SRCNODE |
|
|
79 | in some cases, leading to intra-node messages (such as monitoring) |
|
|
80 | to fail silently (great testcase by Sadrak). |
|
|
81 | |
|
|
82 | 1.27 Sat Mar 20 21:11:16 CET 2010 |
|
|
83 | - avoid exceptions in the transport code if the connection |
|
|
84 | fails during TLS negotiation. |
|
|
85 | - delay new address broadcasts in seednodes, to reduce the |
|
|
86 | likelyhood of nodes connecting at the same time, leading to |
|
|
87 | message loss. |
|
|
88 | - improve general message passing efficiency by 8%, and special |
|
|
89 | case json framing for a 50% speed increase. |
|
|
90 | - add hints how to interface to aemp nodes from other programming |
|
|
91 | languages using the aemp protocol in AnyEvent::MP::Transport |
|
|
92 | (it can really be quite simple). |
|
|
93 | - handle unsupported remote framings more gracefully. |
|
|
94 | |
|
|
95 | 1.26 Wed Dec 30 16:48:37 CET 2009 |
|
|
96 | - INCOMPATIBLE CHANGE: the high-level node protocol changed, |
|
|
97 | all nodes need to upgrade. |
|
|
98 | - remote kil's did not work (reported by Eugene Grigoriev). |
|
|
99 | - documented monitor_timeout, connect_interval, auth_offer, |
|
|
100 | auth_accept, framing_format, autocork and nodelay options |
|
|
101 | and make them configurable via bin/aemp. |
|
|
102 | - some doc updates. |
|
|
103 | - remote kil's were not synchronous to message sends which affected |
|
|
104 | nobody, as remote kils were broken. |
|
|
105 | - remote mon/unmonitor requests were not synchronous to message |
|
|
106 | sends either sometimes, although it likely didnt matter. |
|
|
107 | |
|
|
108 | 1.24 Wed Dec 9 14:59:52 CET 2009 |
|
|
109 | - codename "the foul egg". |
|
|
110 | - increased "unmonitored local port died" message level to 2, so |
|
|
111 | it is displayed by default now. |
|
|
112 | - generate random seed at configure time, not module loading time, |
|
|
113 | so fork&configure works as expected. |
|
|
114 | - decreased "seed"-related message priority in Global from 0 (d'oh) |
|
|
115 | to 9, as they are not signifcant. |
|
|
116 | - new experimental module AnyEvent::MP::DataConn to establish |
|
|
117 | additional data connections between nodes. |
|
|
118 | - bin/aemp now loads AnyEvent::Watchdog::Util before trying |
|
|
119 | to call restart. |
|
|
120 | - bin/aemp setcert didn't properly handle the filename argument. |
|
|
121 | - removed empty and unused timeout= parameter from protocol greeting. |
|
|
122 | |
|
|
123 | 1.23 Fri Nov 6 18:46:26 CET 2009 |
|
|
124 | - listener-less nodes were misinformed by their masters about |
|
|
125 | the location of new nodes, and therefore didn't connect |
|
|
126 | properly in all cases. |
|
|
127 | - messages send during node-up processing could sometimes get lost, |
|
|
128 | which would hamper global's ability to mesh the network. |
|
|
129 | - fixed AnyEvent::MP::Global::grp_reg to return a guard also |
|
|
130 | in scalar context. |
|
|
131 | - fixed AnyEvent::MP::Kernel::mon_nodes to return a guard also |
|
|
132 | in scalar context. |
|
|
133 | - try to improve error reporting when automatically loading |
|
|
134 | a function - being unable to load a module will now |
|
|
135 | stop the process and report the error. |
|
|
136 | |
|
|
137 | 1.22 Sat Oct 17 03:41:47 CEST 2009 |
|
|
138 | - fix two rcv-bugs: after adding a tagged rcv, the default |
|
|
139 | callback would not be used or changed anymore (reported |
|
|
140 | by Felix Antonius Wilhelm Ostmann, with apologies for |
|
|
141 | taking so long to releae the fix). |
|
|
142 | - added Coro::MP introduction to AnyEvent::MP::Intro. |
|
|
143 | - add peval function. |
|
|
144 | |
|
|
145 | 1.21 Mon Oct 5 21:58:36 CEST 2009 |
|
|
146 | - always delay kil messages. |
|
|
147 | - remote monitor requests were never cleaned up (found by |
|
|
148 | Sadrak and analysed by elmex). |
|
|
149 | - move $VERSION to AnyEvent::MP and AnyEvent::MP only. |
|
|
150 | |
|
|
151 | 1.2 Thu Oct 1 22:59:35 CEST 2009 |
|
|
152 | - fix a rather big bug where message processing could be |
|
|
153 | inhibited completely. |
|
|
154 | - grp_mon could not return a guard due to a typo. |
|
|
155 | - implement primitive aemp shell. |
|
|
156 | - do not use bound address as listen length (reported by |
|
|
157 | Felix Antonius Wilhelm Ostman). |
|
|
158 | - no longer export an undefined "reg" function from AE::MP. |
|
|
159 | - the mon_guard return value no longer keeps an additional |
|
|
160 | reference to the passed refs. |
|
|
161 | |
|
|
162 | 1.1 Fri Sep 11 04:34:03 CEST 2009 |
|
|
163 | - bumped the transport protocol to version 1 - sorry, but there |
|
|
164 | were too many bugfixes/changes. |
|
|
165 | - new function AnyEvent::MP::cal for simple rpc. |
| 12 | - renamed AnyEvent::MP::Global functions to grp_reg|get|mon. |
166 | - renamed AnyEvent::MP::Global functions to grp_reg|get|mon. |
|
|
167 | - implemented monitoring for Global groups. |
| 13 | - removed all userspace time-outs from the transport, instead |
168 | - removed all userspace time-outs from the transport, instead |
| 14 | rely on tcp retransmit timeouts and tcp keepalive. |
169 | rely on tcp retransmit timeouts and tcp keepalive. |
| 15 | - spawn now delays spawns on the local node artificially, so |
170 | - spawn now delays spawns on the local node artificially, so |
| 16 | it can return before invoking the init function. |
171 | it can return before invoking the init function. |
|
|
172 | - break endless recursion between ping-pong ports on the local |
|
|
173 | node after ~50 iterations. |
| 17 | - support JSON-encoded aemp arguments. |
174 | - support JSON-encoded aemp arguments. |
| 18 | - added aemp restart. |
175 | - added aemp restart. |
| 19 | - support for starting init functions with parameters added. |
176 | - support for starting init functions with parameters added. |
| 20 | - data_format, auth_offer and auth_accept are now configurable, albeit |
177 | - data_format, auth_offer and auth_accept are now configurable, albeit |
| 21 | not documented. |
178 | not documented. |
|
|
179 | - new service: AnyEvent::MP::LogCatcher. |
|
|
180 | - rely on perl srand'ing better than we could. |
|
|
181 | - try to backport to 5.8.2, unfortunately, recursive dependencies |
|
|
182 | rely on Module::Build, which fails at 5.8.7, so YMMV. |
|
|
183 | - use a vastly different reseeding strategy. |
|
|
184 | - use seed nodes to notify nodes about new nodes. |
|
|
185 | - in the case where a ndoe learns new addresses after it tried |
|
|
186 | connecting, prefer the new addrseses over the old ones. |
| 22 | |
187 | |
| 23 | 1.0 Wed Sep 2 20:40:43 CEST 2009 |
188 | 1.0 Wed Sep 2 20:40:43 CEST 2009 |
| 24 | - randomise more connection times, but reduce delay to 0.2s (seeds) |
189 | - randomise more connection times, but reduce delay to 0.2s (seeds) |
| 25 | and 0.05s (normal connects). |
190 | and 0.05s (normal connects). |
| 26 | - upgrade to and require AnyEvent::Watchdog 1.0. |
191 | - upgrade to and require AnyEvent::Watchdog 1.0. |
