Presence re-entry requirement change for 1.1

[paddybyers]

See ably/docs#711.

After confirming that the lib satisfies the new RTP17c, please update the spec spreadsheet accordingly

┆Issue is synchronized with this Jira Bug by Unito

[ricardopereira]

@paddybyers I'm reviewing this and I think it's updated but maybe I'm missing something. So, on every SYNC message (RTP17c1) the library will do a sync and if the message is the last serial then an automatic re-entry is performed.

Now, the library does a sync when the ATTACHED message has false in HAS_PRESENCE flag (RTP17c2) but it checks some extra checks:

ably-cocoa/Source/ARTRealtimeChannel.m

Lines 698 to 708 in f301785

	if (message.hasPresence) {
	[self.presenceMap startSync];
	}
	else if ([self.presenceMap.members count] > 0 || [self.presenceMap.localMembers count] > 0) {
	if (!message.resumed) {
	// When an ATTACHED message is received without a HAS_PRESENCE flag and PresenceMap has existing members
	[self.presenceMap startSync];
	[self.presenceMap endSync];
	[self.logger debug:__FILE__ line:__LINE__ message:@"R:%p C:%p (%@) PresenceMap has been reset", _realtime, self, self.name];
	}
	}

Is this still valid?
Or the library should do an automatic re-entry every time the HAS_PRESENCE is false?

[ricardopereira]

@paddybyers bump ^

[AndyNicks]

@paddybyers Hey Paddy, can you have a look at Ricardo's comment please

[lukasz-szyszkowski]

As far as I understand it we shouldn't do any additional checks like here:

ably-cocoa/Source/ARTRealtimeChannel.m

Lines 735 to 737 in 48cb1c6

	}
	else if ([self.presenceMap.members count] > 0 || [self.presenceMap.localMembers count] > 0) {
	if (!message.resumed) {

according to RTP17f

The RealtimePresence object should perform automatic re-entry whenever a channel moves into the ATTACHED state. (It does not need to do so for RTL12 ATTACHED events received on already-ATTACHED channels).

Am I correct @paddybyers?

[lawrence-forooghian]

@maratal is currently working on #1494, which updates the library to use protocol version 2 and specification version 2.0.0. This specification version simplifies the criteria for performing an automatic re-entry, replacing RTP17c with RTP17f. I would suggest that we park the current bug and close it once we've implemented RTP17f as part of #1494.

[lawrence-forooghian]

Hey @maratal, would you mind confirming whether you agree with my above assessment, that #1494 will resolve this issue? I'll assign this one to you, to close after #1494. If you disagree with my assessment feel free to assign back to me.

[maratal]

@lawrence-forooghian Correct, RTP17c was replaced with RTP17f and implemented in #1714