[ViewVC] Diff of: cvs/AnyEvent-MP/MP.pm

Comparing AnyEvent-MP/MP.pm (file contents):
Revision 1.1 by root, Thu Jul 30 08:38:50 2009 UTC vs.
Revision 1.69 by root, Sun Aug 30 18:51:49 2009 UTC

…		…
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	use AnyEvent::MP;	7	use AnyEvent::MP;
8		8
		9	$NODE # contains this node's noderef
		10	NODE # returns this node's noderef
		11	NODE $port # returns the noderef of the port
		12
		13	$SELF # receiving/own port id in rcv callbacks
		14
		15	# initialise the node so it can send/receive messages
		16	initialise_node;
		17
		18	# ports are message endpoints
		19
		20	# sending messages
		21	snd $port, type => data...;
		22	snd $port, @msg;
		23	snd @msg_with_first_element_being_a_port;
		24
		25	# creating/using ports, the simple way
		26	my $simple_port = port { my @msg = @_; 0 };
		27
		28	# creating/using ports, tagged message matching
		29	my $port = port;
		30	rcv $port, ping => sub { snd $_[0], "pong"; 0 };
		31	rcv $port, pong => sub { warn "pong received\n"; 0 };
		32
		33	# create a port on another node
		34	my $port = spawn $node, $initfunc, @initdata;
		35
		36	# monitoring
		37	mon $port, $cb->(@msg) # callback is invoked on death
		38	mon $port, $otherport # kill otherport on abnormal death
		39	mon $port, $otherport, @msg # send message on death
		40
		41	=head1 CURRENT STATUS
		42
		43	AnyEvent::MP - stable API, should work
		44	AnyEvent::MP::Intro - outdated
		45	AnyEvent::MP::Kernel - mostly stable
		46	AnyEvent::MP::Global - mostly stable
		47	AnyEvent::MP::Node - mostly stable, but internal anyways
		48	AnyEvent::MP::Transport - mostly stable, but internal anyways
		49
		50	stay tuned.
		51
9	=head1 DESCRIPTION	52	=head1 DESCRIPTION
10		53
		54	This module (-family) implements a simple message passing framework.
		55
		56	Despite its simplicity, you can securely message other processes running
		57	on the same or other hosts, and you can supervise entities remotely.
		58
		59	For an introduction to this module family, see the L<AnyEvent::MP::Intro>
		60	manual page and the examples under F<eg/>.
		61
		62	At the moment, this module family is a bit underdocumented.
		63
		64	=head1 CONCEPTS
		65
		66	=over 4
		67
		68	=item port
		69
		70	A port is something you can send messages to (with the C<snd> function).
		71
		72	Ports allow you to register C<rcv> handlers that can match all or just
		73	some messages. Messages send to ports will not be queued, regardless of
		74	anything was listening for them or not.
		75
		76	=item port ID - C<nodeid#portname>
		77
		78	A port ID is the concatenation of a node ID, a hash-mark (C<#>) as
		79	separator, and a port name (a printable string of unspecified format).
		80
		81	=item node
		82
		83	A node is a single process containing at least one port - the node port,
		84	which enables nodes to manage each other remotely, and to create new
		85	ports.
		86
		87	Nodes are either public (have one or more listening ports) or private
		88	(no listening ports). Private nodes cannot talk to other private nodes
		89	currently.
		90
		91	=item node ID - C<[a-za-Z0-9_\-.:]+>
		92
		93	A node ID is a string that uniquely identifies the node within a
		94	network. Depending on the configuration used, node IDs can look like a
		95	hostname, a hostname and a port, or a random string. AnyEvent::MP itself
		96	doesn't interpret node IDs in any way.
		97
		98	=item binds - C<ip:port>
		99
		100	Nodes can only talk to each other by creating some kind of connection to
		101	each other. To do this, nodes should listen on one or more local transport
		102	endpoints - binds. Currently, only standard C<ip:port> specifications can
		103	be used, which specify TCP ports to listen on.
		104
		105	=item seeds - C<host:port>
		106
		107	When a node starts, it knows nothing about the network. To teach the node
		108	about the network it first has to contact some other node within the
		109	network. This node is called a seed.
		110
		111	Seeds are transport endpoint(s) of as many nodes as one wants. Those nodes
		112	are expected to be long-running, and at least one of those should always
		113	be available. When nodes run out of connections (e.g. due to a network
		114	error), they try to re-establish connections to some seednodes again to
		115	join the network.
		116
		117	Apart from being sued for seeding, seednodes are not special in any way -
		118	every public node can be a seednode.
		119
		120	=back
		121
		122	=head1 VARIABLES/FUNCTIONS
		123
		124	=over 4
		125
11	=cut	126	=cut
12		127
13	package AnyEvent::MP;	128	package AnyEvent::MP;
14		129
		130	use AnyEvent::MP::Kernel;
		131
15	use common::sense;	132	use common::sense;
16		133
		134	use Carp ();
		135
17	use AE ();	136	use AE ();
18		137
19	our $VERSION = '0.0';	138	use base "Exporter";
20		139
21	sub nonce($) {	140	our $VERSION = $AnyEvent::MP::Kernel::VERSION;
22	my $nonce;
23		141
24	if (open my $fh, "</dev/urandom") {	142	our @EXPORT = qw(
25	sysread $fh, $nonce, $_[0];	143	NODE $NODE *SELF node_of after
26	} else {	144	initialise_node
		145	snd rcv mon mon_guard kil reg psub spawn
		146	port
		147	);
		148
		149	our $SELF;
		150
		151	sub _self_die() {
		152	my $msg = $@;
		153	$msg =~ s/\n+$// unless ref $msg;
		154	kil $SELF, die => $msg;
		155	}
		156
		157	=item $thisnode = NODE / $NODE
		158
		159	The C<NODE> function returns, and the C<$NODE> variable contains, the node
		160	ID of the node running in the current process. This value is initialised by
		161	a call to C<initialise_node>.
		162
		163	=item $nodeid = node_of $port
		164
		165	Extracts and returns the node ID from a port ID or a node ID.
		166
		167	=item initialise_node $profile_name, key => value...
		168
		169	Before a node can talk to other nodes on the network (i.e. enter
		170	"distributed mode") it has to initialise itself - the minimum a node needs
		171	to know is its own name, and optionally it should know the addresses of
		172	some other nodes in the network to discover other nodes.
		173
		174	This function initialises a node - it must be called exactly once (or
		175	never) before calling other AnyEvent::MP functions.
		176
		177	The first argument is a profile name. If it is C<undef> or missing, then
		178	the current nodename will be used instead (i.e. F<uname -n>).
		179
		180	The function first looks up the profile in the aemp configuration (see the
		181	L<aemp> commandline utility). the profile is calculated as follows:
		182
		183	First, all remaining key => value pairs will be used. Then they will be
		184	overwritten by any values specified in the global default configuration
		185	(see the F<aemp> utility), then the chain of profiles selected, if
		186	any. That means that the values specified in the profile have highest
		187	priority and the values specified via C<initialise_node> have lowest
		188	priority.
		189
		190	If the profile specifies a node ID, then this will become the node ID of
		191	this process. If not, then the profile name will be used as node ID. The
		192	special node ID of C<anon/> will be replaced by a random node ID.
		193
		194	The next step is to look up the binds in the profile, followed by binding
		195	aemp protocol listeners on all binds specified (it is possible and valid
		196	to have no binds, meaning that the node cannot be contacted form the
		197	outside. This means the node cannot talk to other nodes that also have no
		198	binds, but it can still talk to all "normal" nodes).
		199
		200	If the profile does not specify a binds list, then the node ID will be
		201	treated as if it were of the form C<host:port>, which will be resolved and
		202	used as binds list.
		203
		204	Lastly, the seeds list from the profile is passed to the
		205	L<AnyEvent::MP::Global> module, which will then use it to keep
		206	connectivity with at least on of those seed nodes at any point in time.
		207
		208	Example: become a distributed node listening on the guessed noderef, or
		209	the one specified via C<aemp> for the current node. This should be the
		210	most common form of invocation for "daemon"-type nodes.
		211
		212	initialise_node;
		213
		214	Example: become an anonymous node. This form is often used for commandline
		215	clients.
		216
		217	initialise_node "anon/";
		218
		219	Example: become a distributed node. If there is no profile of the given
		220	name, or no binds list was specified, resolve C<localhost:4044> and bind
		221	on the resulting addresses.
		222
		223	initialise_node "localhost:4044";
		224
		225	=item $SELF
		226
		227	Contains the current port id while executing C<rcv> callbacks or C<psub>
		228	blocks.
		229
		230	=item *SELF, SELF, %SELF, @SELF...
		231
		232	Due to some quirks in how perl exports variables, it is impossible to
		233	just export C<$SELF>, all the symbols named C<SELF> are exported by this
		234	module, but only C<$SELF> is currently used.
		235
		236	=item snd $port, type => @data
		237
		238	=item snd $port, @msg
		239
		240	Send the given message to the given port, which can identify either a
		241	local or a remote port, and must be a port ID.
		242
		243	While the message can be almost anything, it is highly recommended to
		244	use a string as first element (a port ID, or some word that indicates a
		245	request type etc.) and to consist if only simple perl values (scalars,
		246	arrays, hashes) - if you think you need to pass an object, think again.
		247
		248	The message data logically becomes read-only after a call to this
		249	function: modifying any argument (or values referenced by them) is
		250	forbidden, as there can be considerable time between the call to C<snd>
		251	and the time the message is actually being serialised - in fact, it might
		252	never be copied as within the same process it is simply handed to the
		253	receiving port.
		254
		255	The type of data you can transfer depends on the transport protocol: when
		256	JSON is used, then only strings, numbers and arrays and hashes consisting
		257	of those are allowed (no objects). When Storable is used, then anything
		258	that Storable can serialise and deserialise is allowed, and for the local
		259	node, anything can be passed. Best rely only on the common denominator of
		260	these.
		261
		262	=item $local_port = port
		263
		264	Create a new local port object and returns its port ID. Initially it has
		265	no callbacks set and will throw an error when it receives messages.
		266
		267	=item $local_port = port { my @msg = @_ }
		268
		269	Creates a new local port, and returns its ID. Semantically the same as
		270	creating a port and calling C<rcv $port, $callback> on it.
		271
		272	The block will be called for every message received on the port, with the
		273	global variable C<$SELF> set to the port ID. Runtime errors will cause the
		274	port to be C<kil>ed. The message will be passed as-is, no extra argument
		275	(i.e. no port ID) will be passed to the callback.
		276
		277	If you want to stop/destroy the port, simply C<kil> it:
		278
		279	my $port = port {
		280	my @msg = @_;
27	# shit...	281	...
28	our $nonce_init;	282	kil $SELF;
29	unless ($nonce_init++) {
30	srand time ^ $$ ^ unpack "%L*", qx"ps -edalf" . qx"ipconfig /all";
31	}
32
33	$nonce = join "", map +(chr rand 256), 1 .. $_[0]
34	}	283	};
35		284
36	$nonce	285	=cut
37	}
38		286
39	our $DEFAULT_SECRET;	287	sub rcv($@);
40		288
41	sub default_secret {	289	sub _kilme {
42	unless (defined $DEFAULT_SECRET) {	290	die "received message on port without callback";
43	if (open my $fh, "<$ENV{HOME}/.aemp-secret") {	291	}
44	sysread $fh, $DEFAULT_SECRET, -s $fh;	292
		293	sub port(;&) {
		294	my $id = "$UNIQ." . $ID++;
		295	my $port = "$NODE#$id";
		296
		297	rcv $port, shift \|\| \&_kilme;
		298
		299	$port
		300	}
		301
		302	=item rcv $local_port, $callback->(@msg)
		303
		304	Replaces the default callback on the specified port. There is no way to
		305	remove the default callback: use C<sub { }> to disable it, or better
		306	C<kil> the port when it is no longer needed.
		307
		308	The global C<$SELF> (exported by this module) contains C<$port> while
		309	executing the callback. Runtime errors during callback execution will
		310	result in the port being C<kil>ed.
		311
		312	The default callback received all messages not matched by a more specific
		313	C<tag> match.
		314
		315	=item rcv $local_port, tag => $callback->(@msg_without_tag), ...
		316
		317	Register (or replace) callbacks to be called on messages starting with the
		318	given tag on the given port (and return the port), or unregister it (when
		319	C<$callback> is C<$undef> or missing). There can only be one callback
		320	registered for each tag.
		321
		322	The original message will be passed to the callback, after the first
		323	element (the tag) has been removed. The callback will use the same
		324	environment as the default callback (see above).
		325
		326	Example: create a port and bind receivers on it in one go.
		327
		328	my $port = rcv port,
		329	msg1 => sub { ... },
		330	msg2 => sub { ... },
		331	;
		332
		333	Example: create a port, bind receivers and send it in a message elsewhere
		334	in one go:
		335
		336	snd $otherport, reply =>
		337	rcv port,
		338	msg1 => sub { ... },
		339	...
		340	;
		341
		342	Example: temporarily register a rcv callback for a tag matching some port
		343	(e.g. for a rpc reply) and unregister it after a message was received.
		344
		345	rcv $port, $otherport => sub {
		346	my @reply = @_;
		347
		348	rcv $SELF, $otherport;
		349	};
		350
		351	=cut
		352
		353	sub rcv($@) {
		354	my $port = shift;
		355	my ($noderef, $portid) = split /#/, $port, 2;
		356
		357	$NODE{$noderef} == $NODE{""}
		358	or Carp::croak "$port: rcv can only be called on local ports, caught";
		359
		360	while (@_) {
		361	if (ref $_[0]) {
		362	if (my $self = $PORT_DATA{$portid}) {
		363	"AnyEvent::MP::Port" eq ref $self
		364	or Carp::croak "$port: rcv can only be called on message matching ports, caught";
		365
		366	$self->[2] = shift;
45	} else {	367	} else {
46	$DEFAULT_SECRET = nonce 32;	368	my $cb = shift;
		369	$PORT{$portid} = sub {
		370	local $SELF = $port;
		371	eval { &$cb }; _self_die if $@;
		372	};
		373	}
		374	} elsif (defined $_[0]) {
		375	my $self = $PORT_DATA{$portid} \|\|= do {
		376	my $self = bless [$PORT{$port} \|\| sub { }, { }, $port], "AnyEvent::MP::Port";
		377
		378	$PORT{$portid} = sub {
		379	local $SELF = $port;
		380
		381	if (my $cb = $self->[1]{$_[0]}) {
		382	shift;
		383	eval { &$cb }; _self_die if $@;
		384	} else {
		385	&{ $self->[0] };
		386	}
		387	};
		388
		389	$self
		390	};
		391
		392	"AnyEvent::MP::Port" eq ref $self
		393	or Carp::croak "$port: rcv can only be called on message matching ports, caught";
		394
		395	my ($tag, $cb) = splice @_, 0, 2;
		396
		397	if (defined $cb) {
		398	$self->[1]{$tag} = $cb;
		399	} else {
		400	delete $self->[1]{$tag};
		401	}
47	}	402	}
48	}	403	}
49		404
50	$DEFAULT_SECRET	405	$port
51	}	406	}
		407
		408	=item $closure = psub { BLOCK }
		409
		410	Remembers C<$SELF> and creates a closure out of the BLOCK. When the
		411	closure is executed, sets up the environment in the same way as in C<rcv>
		412	callbacks, i.e. runtime errors will cause the port to get C<kil>ed.
		413
		414	This is useful when you register callbacks from C<rcv> callbacks:
		415
		416	rcv delayed_reply => sub {
		417	my ($delay, @reply) = @_;
		418	my $timer = AE::timer $delay, 0, psub {
		419	snd @reply, $SELF;
		420	};
		421	};
		422
		423	=cut
		424
		425	sub psub(&) {
		426	my $cb = shift;
		427
		428	my $port = $SELF
		429	or Carp::croak "psub can only be called from within rcv or psub callbacks, not";
		430
		431	sub {
		432	local $SELF = $port;
		433
		434	if (wantarray) {
		435	my @res = eval { &$cb };
		436	_self_die if $@;
		437	@res
		438	} else {
		439	my $res = eval { &$cb };
		440	_self_die if $@;
		441	$res
		442	}
		443	}
		444	}
		445
		446	=item $guard = mon $port, $cb->(@reason) # call $cb when $port dies
		447
		448	=item $guard = mon $port, $rcvport # kill $rcvport when $port dies
		449
		450	=item $guard = mon $port # kill $SELF when $port dies
		451
		452	=item $guard = mon $port, $rcvport, @msg # send a message when $port dies
		453
		454	Monitor the given port and do something when the port is killed or
		455	messages to it were lost, and optionally return a guard that can be used
		456	to stop monitoring again.
		457
		458	C<mon> effectively guarantees that, in the absence of hardware failures,
		459	after starting the monitor, either all messages sent to the port will
		460	arrive, or the monitoring action will be invoked after possible message
		461	loss has been detected. No messages will be lost "in between" (after
		462	the first lost message no further messages will be received by the
		463	port). After the monitoring action was invoked, further messages might get
		464	delivered again.
		465
		466	Note that monitoring-actions are one-shot: once messages are lost (and a
		467	monitoring alert was raised), they are removed and will not trigger again.
		468
		469	In the first form (callback), the callback is simply called with any
		470	number of C<@reason> elements (no @reason means that the port was deleted
		471	"normally"). Note also that I<< the callback B<must> never die >>, so use
		472	C<eval> if unsure.
		473
		474	In the second form (another port given), the other port (C<$rcvport>)
		475	will be C<kil>'ed with C<@reason>, iff a @reason was specified, i.e. on
		476	"normal" kils nothing happens, while under all other conditions, the other
		477	port is killed with the same reason.
		478
		479	The third form (kill self) is the same as the second form, except that
		480	C<$rvport> defaults to C<$SELF>.
		481
		482	In the last form (message), a message of the form C<@msg, @reason> will be
		483	C<snd>.
		484
		485	As a rule of thumb, monitoring requests should always monitor a port from
		486	a local port (or callback). The reason is that kill messages might get
		487	lost, just like any other message. Another less obvious reason is that
		488	even monitoring requests can get lost (for exmaple, when the connection
		489	to the other node goes down permanently). When monitoring a port locally
		490	these problems do not exist.
		491
		492	Example: call a given callback when C<$port> is killed.
		493
		494	mon $port, sub { warn "port died because of <@_>\n" };
		495
		496	Example: kill ourselves when C<$port> is killed abnormally.
		497
		498	mon $port;
		499
		500	Example: send us a restart message when another C<$port> is killed.
		501
		502	mon $port, $self => "restart";
		503
		504	=cut
		505
		506	sub mon {
		507	my ($noderef, $port) = split /#/, shift, 2;
		508
		509	my $node = $NODE{$noderef} \|\| add_node $noderef;
		510
		511	my $cb = @_ ? shift : $SELF \|\| Carp::croak 'mon: called with one argument only, but $SELF not set,';
		512
		513	unless (ref $cb) {
		514	if (@_) {
		515	# send a kill info message
		516	my (@msg) = ($cb, @_);
		517	$cb = sub { snd @msg, @_ };
		518	} else {
		519	# simply kill other port
		520	my $port = $cb;
		521	$cb = sub { kil $port, @_ if @_ };
		522	}
		523	}
		524
		525	$node->monitor ($port, $cb);
		526
		527	defined wantarray
		528	and AnyEvent::Util::guard { $node->unmonitor ($port, $cb) }
		529	}
		530
		531	=item $guard = mon_guard $port, $ref, $ref...
		532
		533	Monitors the given C<$port> and keeps the passed references. When the port
		534	is killed, the references will be freed.
		535
		536	Optionally returns a guard that will stop the monitoring.
		537
		538	This function is useful when you create e.g. timers or other watchers and
		539	want to free them when the port gets killed (note the use of C<psub>):
		540
		541	$port->rcv (start => sub {
		542	my $timer; $timer = mon_guard $port, AE::timer 1, 1, psub {
		543	undef $timer if 0.9 < rand;
		544	});
		545	});
		546
		547	=cut
		548
		549	sub mon_guard {
		550	my ($port, @refs) = @_;
		551
		552	#TODO: mon-less form?
		553
		554	mon $port, sub { 0 && @refs }
		555	}
		556
		557	=item kil $port[, @reason]
		558
		559	Kill the specified port with the given C<@reason>.
		560
		561	If no C<@reason> is specified, then the port is killed "normally" (ports
		562	monitoring other ports will not necessarily die because a port dies
		563	"normally").
		564
		565	Otherwise, linked ports get killed with the same reason (second form of
		566	C<mon>, see above).
		567
		568	Runtime errors while evaluating C<rcv> callbacks or inside C<psub> blocks
		569	will be reported as reason C<< die => $@ >>.
		570
		571	Transport/communication errors are reported as C<< transport_error =>
		572	$message >>.
		573
		574	=cut
		575
		576	=item $port = spawn $node, $initfunc[, @initdata]
		577
		578	Creates a port on the node C<$node> (which can also be a port ID, in which
		579	case it's the node where that port resides).
		580
		581	The port ID of the newly created port is returned immediately, and it is
		582	possible to immediately start sending messages or to monitor the port.
		583
		584	After the port has been created, the init function is called on the remote
		585	node, in the same context as a C<rcv> callback. This function must be a
		586	fully-qualified function name (e.g. C<MyApp::Chat::Server::init>). To
		587	specify a function in the main program, use C<::name>.
		588
		589	If the function doesn't exist, then the node tries to C<require>
		590	the package, then the package above the package and so on (e.g.
		591	C<MyApp::Chat::Server>, C<MyApp::Chat>, C<MyApp>) until the function
		592	exists or it runs out of package names.
		593
		594	The init function is then called with the newly-created port as context
		595	object (C<$SELF>) and the C<@initdata> values as arguments.
		596
		597	A common idiom is to pass a local port, immediately monitor the spawned
		598	port, and in the remote init function, immediately monitor the passed
		599	local port. This two-way monitoring ensures that both ports get cleaned up
		600	when there is a problem.
		601
		602	Example: spawn a chat server port on C<$othernode>.
		603
		604	# this node, executed from within a port context:
		605	my $server = spawn $othernode, "MyApp::Chat::Server::connect", $SELF;
		606	mon $server;
		607
		608	# init function on C<$othernode>
		609	sub connect {
		610	my ($srcport) = @_;
		611
		612	mon $srcport;
		613
		614	rcv $SELF, sub {
		615	...
		616	};
		617	}
		618
		619	=cut
		620
		621	sub _spawn {
		622	my $port = shift;
		623	my $init = shift;
		624
		625	local $SELF = "$NODE#$port";
		626	eval {
		627	&{ load_func $init }
		628	};
		629	_self_die if $@;
		630	}
		631
		632	sub spawn(@) {
		633	my ($noderef, undef) = split /#/, shift, 2;
		634
		635	my $id = "$RUNIQ." . $ID++;
		636
		637	$_[0] =~ /::/
		638	or Carp::croak "spawn init function must be a fully-qualified name, caught";
		639
		640	snd_to_func $noderef, "AnyEvent::MP::_spawn" => $id, @_;
		641
		642	"$noderef#$id"
		643	}
		644
		645	=item after $timeout, @msg
		646
		647	=item after $timeout, $callback
		648
		649	Either sends the given message, or call the given callback, after the
		650	specified number of seconds.
		651
		652	This is simply a utility function that comes in handy at times - the
		653	AnyEvent::MP author is not convinced of the wisdom of having it, though,
		654	so it may go away in the future.
		655
		656	=cut
		657
		658	sub after($@) {
		659	my ($timeout, @action) = @_;
		660
		661	my $t; $t = AE::timer $timeout, 0, sub {
		662	undef $t;
		663	ref $action[0]
		664	? $action[0]()
		665	: snd @action;
		666	};
		667	}
		668
		669	=back
		670
		671	=head1 AnyEvent::MP vs. Distributed Erlang
		672
		673	AnyEvent::MP got lots of its ideas from distributed Erlang (Erlang node
		674	== aemp node, Erlang process == aemp port), so many of the documents and
		675	programming techniques employed by Erlang apply to AnyEvent::MP. Here is a
		676	sample:
		677
		678	http://www.Erlang.se/doc/programming_rules.shtml
		679	http://Erlang.org/doc/getting_started/part_frame.html # chapters 3 and 4
		680	http://Erlang.org/download/Erlang-book-part1.pdf # chapters 5 and 6
		681	http://Erlang.org/download/armstrong_thesis_2003.pdf # chapters 4 and 5
		682
		683	Despite the similarities, there are also some important differences:
		684
		685	=over 4
		686
		687	=item * Node IDs are arbitrary strings in AEMP.
		688
		689	Erlang relies on special naming and DNS to work everywhere in the same
		690	way. AEMP relies on each node somehow knowing its own address(es) (e.g. by
		691	configuraiton or DNS), but will otherwise discover other odes itself.
		692
		693	=item * Erlang has a "remote ports are like local ports" philosophy, AEMP
		694	uses "local ports are like remote ports".
		695
		696	The failure modes for local ports are quite different (runtime errors
		697	only) then for remote ports - when a local port dies, you I<know> it dies,
		698	when a connection to another node dies, you know nothing about the other
		699	port.
		700
		701	Erlang pretends remote ports are as reliable as local ports, even when
		702	they are not.
		703
		704	AEMP encourages a "treat remote ports differently" philosophy, with local
		705	ports being the special case/exception, where transport errors cannot
		706	occur.
		707
		708	=item * Erlang uses processes and a mailbox, AEMP does not queue.
		709
		710	Erlang uses processes that selectively receive messages, and therefore
		711	needs a queue. AEMP is event based, queuing messages would serve no
		712	useful purpose. For the same reason the pattern-matching abilities of
		713	AnyEvent::MP are more limited, as there is little need to be able to
		714	filter messages without dequeing them.
		715
		716	(But see L<Coro::MP> for a more Erlang-like process model on top of AEMP).
		717
		718	=item * Erlang sends are synchronous, AEMP sends are asynchronous.
		719
		720	Sending messages in Erlang is synchronous and blocks the process (and
		721	so does not need a queue that can overflow). AEMP sends are immediate,
		722	connection establishment is handled in the background.
		723
		724	=item * Erlang suffers from silent message loss, AEMP does not.
		725
		726	Erlang makes few guarantees on messages delivery - messages can get lost
		727	without any of the processes realising it (i.e. you send messages a, b,
		728	and c, and the other side only receives messages a and c).
		729
		730	AEMP guarantees correct ordering, and the guarantee that after one message
		731	is lost, all following ones sent to the same port are lost as well, until
		732	monitoring raises an error, so there are no silent "holes" in the message
		733	sequence.
		734
		735	=item * Erlang can send messages to the wrong port, AEMP does not.
		736
		737	In Erlang it is quite likely that a node that restarts reuses a process ID
		738	known to other nodes for a completely different process, causing messages
		739	destined for that process to end up in an unrelated process.
		740
		741	AEMP never reuses port IDs, so old messages or old port IDs floating
		742	around in the network will not be sent to an unrelated port.
		743
		744	=item * Erlang uses unprotected connections, AEMP uses secure
		745	authentication and can use TLS.
		746
		747	AEMP can use a proven protocol - TLS - to protect connections and
		748	securely authenticate nodes.
		749
		750	=item * The AEMP protocol is optimised for both text-based and binary
		751	communications.
		752
		753	The AEMP protocol, unlike the Erlang protocol, supports both programming
		754	language independent text-only protocols (good for debugging) and binary,
		755	language-specific serialisers (e.g. Storable). By default, unless TLS is
		756	used, the protocol is actually completely text-based.
		757
		758	It has also been carefully designed to be implementable in other languages
		759	with a minimum of work while gracefully degrading functionality to make the
		760	protocol simple.
		761
		762	=item * AEMP has more flexible monitoring options than Erlang.
		763
		764	In Erlang, you can chose to receive I<all> exit signals as messages
		765	or I<none>, there is no in-between, so monitoring single processes is
		766	difficult to implement. Monitoring in AEMP is more flexible than in
		767	Erlang, as one can choose between automatic kill, exit message or callback
		768	on a per-process basis.
		769
		770	=item * Erlang tries to hide remote/local connections, AEMP does not.
		771
		772	Monitoring in Erlang is not an indicator of process death/crashes, in the
		773	same way as linking is (except linking is unreliable in Erlang).
		774
		775	In AEMP, you don't "look up" registered port names or send to named ports
		776	that might or might not be persistent. Instead, you normally spawn a port
		777	on the remote node. The init function monitors you, and you monitor the
		778	remote port. Since both monitors are local to the node, they are much more
		779	reliable (no need for C<spawn_link>).
		780
		781	This also saves round-trips and avoids sending messages to the wrong port
		782	(hard to do in Erlang).
		783
		784	=back
		785
		786	=head1 RATIONALE
		787
		788	=over 4
		789
		790	=item Why strings for port and node IDs, why not objects?
		791
		792	We considered "objects", but found that the actual number of methods
		793	that can be called are quite low. Since port and node IDs travel over
		794	the network frequently, the serialising/deserialising would add lots of
		795	overhead, as well as having to keep a proxy object everywhere.
		796
		797	Strings can easily be printed, easily serialised etc. and need no special
		798	procedures to be "valid".
		799
		800	And as a result, a miniport consists of a single closure stored in a
		801	global hash - it can't become much cheaper.
		802
		803	=item Why favour JSON, why not a real serialising format such as Storable?
		804
		805	In fact, any AnyEvent::MP node will happily accept Storable as framing
		806	format, but currently there is no way to make a node use Storable by
		807	default (although all nodes will accept it).
		808
		809	The default framing protocol is JSON because a) JSON::XS is many times
		810	faster for small messages and b) most importantly, after years of
		811	experience we found that object serialisation is causing more problems
		812	than it solves: Just like function calls, objects simply do not travel
		813	easily over the network, mostly because they will always be a copy, so you
		814	always have to re-think your design.
		815
		816	Keeping your messages simple, concentrating on data structures rather than
		817	objects, will keep your messages clean, tidy and efficient.
		818
		819	=back
52		820
53	=head1 SEE ALSO	821	=head1 SEE ALSO
		822
		823	L<AnyEvent::MP::Intro> - a gentle introduction.
		824
		825	L<AnyEvent::MP::Kernel> - more, lower-level, stuff.
		826
		827	L<AnyEvent::MP::Global> - network maintainance and port groups, to find
		828	your applications.
54		829
55	L<AnyEvent>.	830	L<AnyEvent>.
56		831
57	=head1 AUTHOR	832	=head1 AUTHOR
58		833

Diff Legend

-–
+Removed lines
-+
+Added lines
-<
+Changed lines
->
+Changed lines

Comparing AnyEvent-MP/MP.pm (file contents): Revision 1.1 by root, Thu Jul 30 08:38:50 2009 UTC vs. Revision 1.69 by root, Sun Aug 30 18:51:49 2009 UTC

Diff Legend

Comparing AnyEvent-MP/MP.pm (file contents):
Revision 1.1 by root, Thu Jul 30 08:38:50 2009 UTC vs.
Revision 1.69 by root, Sun Aug 30 18:51:49 2009 UTC