Tuesday, May 12, 2015

An in-principle solution for robust device agnostic engagement.

I think I have been pondering this for several years since I dreamt up the "subscriber" mechanism as being a solution that will facilitate the establishment of "synapse" channels between Clique Space aware devices.

Today, I formalised what I think is a robust mechanism that uses structures developed over this time. The mechanism was seen through murky waters until I had finally developed the necessary data structures to carry it.

It is summed up by the diagram below.


The diagram shows two Clique Space aware devices (Client Devices - I'm beginning to saunter up to an earlier nomenclature that seems to have recently become more appealing, but possibly more on that in a later post).

The initiator, labelled as the upper rectangle and the respondent, labelled as the lower rectangle. The initiator can be any Agent Device. The respondent can be any Client Device (and Clique Space aware device that can participate in a viscera, including Agent Devices and administrator clients).

Furthermore, the diagram shows how these Client Devices engage; how they establish their synapses; how they establish and communicate the Participants within each synapse.

The diagram can be further summed up by this slightly more formalised list, indicating which Participant is created at each stage of the process. The numbers coincide with each of the circled numbers in the diagram above.
  1. Initiator : Owner : Inbound : Create.
  2. Respondent : Owner : Outbound : Reconstituted from 1 as parameter.
  3. Respondent : Owner : Inbound : Create.
  4. Initiator : Owner : Outbound : Reconstituted from 3 as parameter.
  5. Initiator : Non-Owner : Outbound : Create.
  6. Respondent : Non-Owner : Inbound : Reconstituted from 5 as return.
  7. Respondent : Non-Owner : Outbound : Create.
  8. Initiator : Non-Owner : Inbound : Reconstituted from 7 as return.
The Owner/Non-Owner distinction signifies the role that a given participant plays in the synapse being modelled in a bipartite Clique. The Clique is bipartite because it is modelling an engagement between two Client Devices.

The Inbound/Outbound distinction signifies the perspective each device has to the two channels. If, for example, a Client Device A is engaged with Client Device B, then the Inbound channel on A is the Outbound channel on device B. In this way, the synapse Cliques model directionality of both synapses between the two devices.

In all instances where the two engaging devices are both Agent Devices, Participants which are created by a device are possessed by the device. In instances where an Agent Device is engaging an administrator client (I'm thinking about calling this an administrator device), all Participants are possessed by the Agent Device.

So, the device initiating the engagement (possibly, but not exclusively through the engager server's "engage" method, firstly creates the Owner of its inbound channel. The process that completes an engagement as a respondent is then called by the initiating device on the responding device where the respondent first reconstitutes the owner of its outbound channel using the signal given to it by the initiator, and then creates the owner of its outbound channel. Next, the respondent asks the initiator to complete its engagement. The initiator first reconstitutes the owner of the outbound channel from the signal given to it by the respondent, and then creates the non-owner of the same channel. When the engager finishes, the respondent creates the non-owner of its inbound channel using the signal that is returned, and creates a non-owner for its outbound channel before the respondent process completes whereby control is returned back to the initiator. Finally, the initiator reconstitutes the inbound channel's non-owner using the signal that is returned.

Now, to implement...

Friday, April 24, 2015

Robust disengagemt? Hmmm...

As of about one and a half hours ago, I've been able to disengage one administrator client from one Agent Device. The question now is: is this operation sufficiently generic to disengage more than one administrator client from one Agent Device, and furthermore, is this operation sufficiently generic to disengage two Agent Devices.

The first answer is: yes! It appears that an Agent Device can engage and disengage multiple administrator clients. It does not appear to matter whether these engagements are simultaneous. Additionally, one administrator client is able to simultaneously engage and disengage with more than one Agent Device. All of this is an intention designed from conception in mid-2004 to make Clique Space(TM) useful.

The larger question of engaging two Agent Devices should be answered in the next revision. It is hoped that once two Agent Devices can robustly engage and disengage, then the proposition of constructing neural clusters of Agent Devices might be seriously considered.

Wednesday, April 22, 2015

It works?

A Clique Space(TM)  device appears almost to function as a quasi-autonomous agent. There is, however, one obstacle: the disengagement operation.

I've discussed this subject before in past blog entries. Any two devices that engage must necessarily disengage, but disengagement is not engagement in reverse. Instead, the process by which two devices engage each other is a process whereby structures that synchronise both (two synapses, two corresponding postsynaptic subscribers) are created; disengagement is a process where these structures are destroyed. There is no symmetry beyond this obvious necessity.

In the creation of a synapse, the two devices must synchronise the state of each other's postsynaptic subscribers created for this engagment, and this synchronisation must continue until both devices have disengaged whereupon these postsynaptic subscribers are destroyed.

I believe I have a fairly robust solution for engagement. Although the same cannot be said of the disengagement, maybe I am almost there...

Monday, February 16, 2015

Generic beauty.

I think I did something today that I might just label the "end of the beginning" for development activity on the CliqueSpace(TM) core implementation.

Something appeared finally to vault the generic structure: I discarded the notion of a "root" profile. Instead, I made the Participant implement the Media and Mode Profile's interfaces.

Many problems appeared to dissolve... perhaps.

Sunday, January 25, 2015

It Works! Part 1.

Finally, after 6.5 years of coding, Clique Space(TM) works... kind of.

But that's right... it does indeed do something non-trivial. An Agent Device will now engage and disengage one or more administrator clients. I just got it working... not more than 20 minutes ago.

The synapse: works.
The service: works.
The subscriber: works.
The subscription: works.
The context: works.
The payload: works.
The transmitter: works.
The device: works.
The signal: works.
The Element: works.
The principle: works.
The root profile: works.
The Participant: works.

These components comprise most of the implementation of Clique Space's real-time manifestation mechanism. This mechanism (for Agent Device - administrator client engagements only): works.

Now, I've gotta get two or more Agent Devices to engage themselves over a Sovereign's Clique Space. That is, I've got to get Agent Devices to form neural clusters based on a secret that is never shared between devices called "That Which is Sacred"(TM... perhaps?) - the Clique Spaces that manifest Sovereigns, modelled in a Clique(TM) called the Viscera(TM).

After that, I'll:
  1. try and get two or more Agent Devices from different Sovereigns to engage each other in a "Congressional(TM) (formerly known as an Agent Collaboration) Clique Space.
  2. try to make devices more robust and able to reconfigure engagements in adverse environmental conditions.

I think that the real-time reconfiguration under adversity is really going to take some time to nut out.

To the viscera and beyond...

.. gotta love all this TM'ing. Well, it is mine, so mits off!

Tuesday, December 9, 2014

CL,I,P,O,N

A nightmare of confusion.

Recent time with Clique Space(TM) has dragged me into the dependency nightmare that is Java generics; a nightmare from which I might be awakening.

Clique Space code comes in two halves. The chassis is the basic Clique Space data structure definitions and implementation of functionality in relation to the sending and receiving of transmitters while the top half is the body of the specific implementation of a Clique Space aware device.

The halves are separated by a set of parameterised types that make the cute acronym disclosed in the title of this blog entry. The specific device implementation's body "clips on" to the generic chassis. Hence, the chassis defines these generic parameters, and the body completes their implementation. The terms used in the acronym are as follows:
  • CL - The basic Clique structure and mechanisms contained therein.
  • I - The Identity structure and mechanisms contained therein.
  • P - The basic Participant structure and mechanisms contained therein.
  • O - A more specific Participant structure relating to Owner Participants of synapses and involvements.
  • N - A more specific Participant structure relating to non-Owner Participants of synapses and involvements.
The acronym using the first three of these parameterisations has been around for a while. The final two types needed to be added to or "clipped on" the end of the fist three because I wanted a way to discriminate between the Owner and non-Owner Participants of a synapse or a service or viscera as involvements. One of the prime benefits of using generics is that one can check types at compile time, thereby reducing the size of one's code (not having to guard run time type casting) and ensuring one's code has fewer bugs.

Hence, these extra two types (O and N) are used to ensure that an Owner Participant is returned where an Owner Particpant is expected and a non-Owner Participant is likewise returned when expected. This type of discrimination is important in cases where structures derived from the basic Clique (the synapse, the service, and the viscera) are being manipulated. These specific Clique structures have extra functionality because this extra functionality helps a Clique space do the things I intend it to do.

A thing of beauty.

Wednesday, December 3, 2014

The Viscera and the Service

I explained the function of the viscera here, so I will dispense another explanation in this entry and explain what the service is. I'll then contrast the two.

The service, like the viscera, is a Clique. Unlike the viscera, it is strictly a bipartite Clique; the owner is always an Agent Device, and the other member is an external device (which could be a View device like an administrator client which is able to view or persist Clique Space activity). External devices cannot participate in viscera because they are not Agent Devices. Agent Devices are the only type of Clique Space aware device that manifest the functionality necessary to participate in viscera; Agent Devices are the neurons of a Clique Space.

Although services are not synapses (synapses are another type of bipartite Clique) they are used where synapses connect an Agent Device with a View device. A set of synapses do not connect an Agent Device to any other type of external device other than a View device. Hence, external devices that lack a View capability are known simply as drones.

A service implements a connection between an Agent Device and some type of external device, whether that device be a View device or a drone. All external devices siphon their state into a Clique Space and accept commands from a Clique Space through one or more Agent Devices through which each device shares a service.

The basic service represents the aggregate mechanism of exchange between an Agent Device and an external device that exists beyond the synapse mechanism that Clique Space aware devices use to share information. The service may define an interface known as a service delegate which defines how external device information is exchanged between the external device and its connected Agent Device. Services are strictly bipartite because they represent the mechanism of exchange between one external device instance and the Agent Device instance that serves it.

An Agent Device can stop serving a drone by simply disbanding the associated service, while more specifically for a View device, the connected Agent Device must also disband the associated inbound and outbound synapses.

A service is not a viscera. To lightly go over that which has already been explained, a viscera is a Clique that represents all the Agent Devices that contribute to the manifestation of a Clique Space. A viscera is multipartite, and can theoretically contain an unlimited number of participants.

However, a service is a viscera in a subtle but crucial way. Just as every Agent Device obtains a non owner visceral participant to be a member of a viscera, an external device must obtain a non owner service Participant to be counted as a member of a service. This symmetry is very handy because it allows many of the mechanisms of Agent Device - Agent Device engagement to be used in Agent Device - View device engagement. The convenience of this symmetry also appears to confirm that I remain onto a real possibility with this Clique Space thing.