Architecture of morphogenesis

Chapter 5. Architecture of morphogenesis

Hierarchy ROOT → GATE → ANOD · Three-section architecture of ANOD · Splitting of MOVE · Life cycle of an organ · Boundary between MOGE and GNET

5. 1. Hierarchy ROOT → GATE → ANOD

The previous chapters described morphogenesis as a process: what GERM is, how MOVE is arranged, which steps make up RTR0. Open remained the question of the architectural organization in which this process takes place. Where is GERM placed? Who receives the fragments of MOVE and instantiates classes? How are resources distributed among functional organs? This chapter answers these questions.

The architecture of morphogenesis in Gativus is organized as a three-level hierarchy of delegation. Each level has a single, clearly defined role and does not duplicate the functions of the others.

The hierarchy reflects the division of responsibility by management scales. ROOT is responsible for the global integrity of the network — the name space and the basic distribution of authorities. GATE is responsible for the local platform — the reception of the seed, the distribution of its resources among the organs of one organism. ANOD is responsible for one organ — the detailed morphogenesis, the management of that organ's nodes, its life cycle.

The principle of non-extension of functions is essential. ROOT does not manage morphogenesis, does not contain MOVE, does not know which GERMs are placed on which GATEs. GATE does not manage the detailed morphogenesis of concrete organs. ANOD does not interact directly with ROOT and does not extend beyond its own organ. Each level does only its own — and this limitation of functions ensures the scalability of the system.

5. 2. ROOT as root registrar

ROOT — the root node of the Gativus network, performing the function of root registrar. Its functions in morphogenesis are limited and in principle not extended: allocation of the name space and delegation of the basic AVEC₀.

a) Metaphor of the womb

ROOT is the womb of Gativus. The biological womb provides the environment for the development of the embryo: space, nourishment, protection. But the womb does not contain the DNA of the embryo and does not manage its development. It provides conditions, and in those conditions the embryo develops autonomously according to its own genetic program.

ROOT plays an analogous role for Gativus. It provides the basic conditions for the morphogenesis of any organism: a unique UNON name space, an initial AVEC with the rights to create nodes and establish relations. But ROOT does not contain the MOVE of concrete organisms, does not manage their morphogenesis, does not know the details of their structure. These functions are delegated down the hierarchy.

This metaphor has principled architectural significance. When the network scales to a billion GATEs, ROOT can neither manage every germination nor even know about every one of them. It registers names and allocates quotas — and that is sufficient. All other functions are localized.

b) What ROOT does in morphogenesis

The concrete functions of ROOT in the context of morphogenesis reduce to several operations.

Registration of GATE. At the initialization of a new platform, ROOT assigns it a unique 32-bit GATE number and allocates a range of UNON for the nodes that will be created on that platform. This is a one-time operation performed when a GATE joins the network.

Delegation of the initial AVEC₀. ROOT issues a package of assets in the consistency space CNST₀ (see GNET, Chapter 4) and transfers it to the new GATE. This package contains coordinates authorizing the GATE to create a certain number of nodes and establish a certain number of relations. The size of the package is determined by ROOT policy.

Storage of D-components. The ROOT repository contains a library of D-components — trained weights of GTR0 (see Chapter 1, section 1.6). Any GATE platform at morphogenesis establishes d-relations to D-components in ROOT and through these relations obtains the executable code of node classes.

Issuance of global assets. System assets that act globally across the network are issued by ROOT. In the context of morphogenesis this includes NRGN quotas, SYGN quotas, legitimacy markers (see GNET, Chapter 4, section 4.7).

c) What ROOT does not do

The list of functions not performed by ROOT is no less important for understanding the architecture.

ROOT does not store the MOVE of concrete organisms. MOVE lives in the M-section of those nodes that will be unfolded, and if necessary — as assets in the distributed registry (see Chapter 3). ROOT does not intervene in this.

ROOT does not know which GERMs are placed on which GATEs. This is local information of the GATE, not subject to global accounting.

ROOT does not manage morphogenesis. After the delegation of AVEC₀ to GATE, further operations — reception of GERM, splitting of MOVE, neurogenesis, synaptogenesis — are performed autonomously without appeal to ROOT.

ROOT does not receive reports on the progress of morphogenesis. If germination is successful — ROOT is not notified. If unsuccessful — likewise. Local events are not routed up.

d) Availability and sustainability

When ROOT is unavailable, the network continues to work within the previously received AVEC₀. Existing GATEs continue to accept GERMs, ANODs continue morphogenesis, nodes continue to exchange datagrams — all within already delegated quotas. Only operations requiring fresh ROOT issuance are impossible: creating new GATEs, increasing the AVEC₀ of existing ones, resolving global name conflicts.

This is consistent with the principle of the womb: the womb provides conditions, but an embryo that has already begun developing continues to develop autonomously even when the external environment changes. Analogously, a Gativus organism that has passed initial germination does not depend on the current availability of ROOT.

5. 3. GATE as platform of germination

GATE (GATivus Edge) — the physical device on which morphogenesis takes place. It is an edge node of the Gativus network, connecting the computational substrate (processors, memory, disks, network interfaces) with the Gativus architecture. One GATE is one physical or virtual machine with allocated resources. A multi-platform Gativus network consists of many GATEs connected by standard IPv6 protocols.

a) Three functions of GATE in morphogenesis

GATE performs three different functions in morphogenesis, forming a single cycle.

Reception of GERM. GERM arrives at a GATE in one of several ways: loading by the operator at platform initialization, receipt from another GATE through the network, copying from a repository. The received GERM is placed in the M-section of one of the GATE's nodes and awaits the start of germination.

Supplementing AVEC with hardware resources. The basic AVEC₀ received from ROOT contains logical quotas — rights to create nodes and relations. GATE supplements it with physical resources: memory (how many bytes are available for nodes), computation (how many processor cycles per unit time), network bandwidth, disk space. These resources are the property of the GATE, and ROOT neither knows them nor is obliged to know them.

Distribution of resources among ANODs. For every functional organ described in the COMP diagram of MOVE, GATE creates a node ANOD and passes to it a fragment of MOVE and a share of the combined AVEC. This is the central operation of GATE in morphogenesis — discussed in more detail in section 5.5.

b) Several GERMs on one GATE

One GATE can simultaneously contain several GERMs, each starting independent germination within its own share of AVEC. This gives several independent organisms on one platform — an analogue of multiple pregnancy in biology.

Each GERM receives an allocated share of the GATE's physical resources. This division is strict: memory allocated to one GERM is unavailable to others; processor time allocated to one is not used by another. The strict division ensures isolation: problems with the morphogenesis of one GERM do not affect others.

The distribution of resources among GERMs on one GATE is the decision of the platform administrator and is not automatically determined. One GATE can allocate a larger share to one 'priority' GERM and smaller shares to the others; can divide resources equally; can leave part of the resources in reserve. This policy is outside MOGE and belongs to platform operations.

c) GATE as trust boundary

GATE is a natural trust boundary in the Gativus network. Inside one GATE, nodes interact via local fast channels (the G1 layer in GORP, see GNET), which do not require cryptographic protection at the level of each datagram. Between different GATEs, transfer passes through standard internet channels (the G3 layer) with encryption and signature verification.

This boundary is consistent with the physical reality: one GATE is one physical or virtual machine under unified administrative control. Different GATEs are different machines, potentially in different jurisdictions, under different administrative control. Trust between them is not automatically assumed.

5. 4. ANOD as organ administrator

ANOD (Alliance NODe) — the node coordinating the morphogenesis of one functional organ. ANOD is not a dedicated separate entity with its own ontology; it is an ordinary NDDI node with a special set of sections, performing administrative functions for a group of connected nodes.

Each functional organ of Gativus has exactly one ANOD. Inside ANOD the entire morphogenesis of that organ takes place: parsing of the MOVE fragment, neurogenesis, synaptogenesis, maturation, transition to operation. All these processes are autonomous — ANOD does not appeal to GATE or ROOT for permission for each action; it acts within the delegated AVEC and the fragment of MOVE.

a) Three-section architecture

Inside ANOD, three sections are logically distinguished, each responsible for one aspect of the organ's life cycle. This division is principled, ensuring a clear distribution of responsibility and excluding conflicts of functions.

The G-section is responsible for security and integrity. Its contents: RULE matrices (see GNET), AVEC quotas, audit log. Any request for neurogenesis or synaptogenesis passes through the G-section for validation — are quotas sufficient, is the operation permitted, is integrity not violated. The G-section does not build or design — it controls. Biological analogy: the immune system of the organ.

The D-section (also LOAI, Local Asset Issuer) is responsible for the code. Its contents: D-components — object code of node types with their connectors and CONT types. At neurogenesis, LOAI provides code for the assembly of a new node of the requested type. LOAI does not decide what to build and does not initiate building — it provides material. Biological analogy: the gene repository of the organ.

The M-section (also LOMN, Local Morphogenesis Node) is responsible for growth. Its contents: the fragment of MOVE for the concrete organ, the queue of morphogenetic tasks, coordination of processes. LOMN is the subject of morphogenesis. It reads MOVE, requests D-components from LOAI, passes validation in the G-section, launches NRGN, coordinates SYGD and SYGA, controls maturation.

b) Principled nature of the division

The division into three sections is not an abstraction or a decorative device. It is a division of responsibility that excludes conflicts of functions.

LOMN cannot bypass the G-section's quotas. Any request to create a node or establish a deterministic connection passes through the G-section for validation. If the G-section refuses — the operation is not performed. LOMN has no way to create a node bypassing the G-section.

The G-section cannot change D-components. It works with rights and quotas, but not with executable code. If the behaviour of nodes must be changed — this is an update of the D-component in ROOT and its subsequent propagation through d-relations, not a modification of the G-section.

LOAI cannot initiate neurogenesis. It is passive: it provides code on request, but does not itself decide when and how many nodes to create. That is the function of LOMN.

Each section does only its own. This division gives architectural cleanliness and facilitates debugging: if something does not work, diagnosis begins with determining in which section exactly the problem arose.

c) All sections — inside one NDDI

Despite the logical division into three sections, ANOD is one NDDI node, not three. All sections work inside one process, without network overhead between them. The exchange between the G, D, and M sections of ANOD goes through the node's internal interfaces, instantaneously and without the overhead of network transmission.

External communication of ANOD with other nodes — with the parent ANOD, with GATE, with subordinate nodes of the organ — is performed through standard g-relations with multiplexing by message type. From the outside, ANOD looks like one node with one UNON; the internal division into sections is a detail of its own organization.

5. 5. Splitting of MOVE upon delegation

Upon delegation from GATE to ANOD, the MOVE vector is not copied as a whole but split. The full MOVE contains descriptions of all functional organs of the organism; GATE passes to each ANOD only the fragment relating to its own organ. The full MOVE remains with GATE.

a) Mechanics of the splitting

The splitting happens in a single pass. GATE distributes tasks to all ANODs simultaneously: for each component from the COMP diagram, an ANOD is created, and into the M-section of that ANOD the corresponding fragment of MOVE is written.

The MOVE fragment for one ANOD contains:

The complete CLSS diagram of the classes that must be created in this organ. These classes are the property of this ANOD; no other ANOD creates them.

A fragment of the COMM diagram describing the internal connections of the organ — those templates connecting classes inside one component from COMP. Inter-organ connections are described separately (see below).

A fragment of the ACTD diagram describing the morphogenesis procedures for this organ — the order of class instantiation, control points, transitions between phases.

The corresponding share of SPCE — where exactly in the physical space of GATE the nodes of this organ will be located.

The corresponding share of RSRC — what AVEC has been allocated to this organ.

This splitting is possible because Gativus models only subjective reality — the number of functional organs is measured in dozens, not in billions. For one GATE, splitting MOVE among about two dozen ANODs is a feasible operation that creates no bottleneck.

b) Local irreversibility

The splitting is locally irreversible. An ANOD that has received a MOVE fragment for one organ cannot receive a fragment for another organ. Its M-section contains only what relates to its own functional organ, and extension into foreign territory is not envisaged.

This restriction gives architectural cleanliness. Each ANOD knows only its own; there is no possibility of accidentally or maliciously intervening in the morphogenesis of a foreign organ. The biological analogy is direct: a neuron of the visual cortex cannot 'change its mind' and become a neuron of the hippocampus — its fate is determined by the fragment of the genetic program available to it at the moment of differentiation.

c) Global recoverability

Globally, however, the situation is different. GATE has preserved the complete MOVE and can at any moment create a new ANOD with any fragment. If, for example, an existing ANOD has been damaged or lost, GATE can create its replacement — a new ANOD with the same MOVE fragment.

This is not recovery of the damaged ANOD but its replacement. The new ANOD will start morphogenesis anew, and the resulting organ will be a new instance, structurally corresponding to the same MOVE fragment but with different UNONs of the nodes and different concrete topology of SYGA connections. Biological analogy: a stem cell capable of reproducing lost tissue, but giving new tissue, not the original one.

d) Transfer of information about inter-organ connections

An essential aspect of the splitting is inter-organ connections. If a node of organ A must be connected with a node of organ B, this connection cannot be fully described within one MOVE fragment. Class A is defined in ANOD-A, class B in ANOD-B; neither alone has the complete picture.

The solution is in the transfer of CONT types of connectors. When MOVE is split, each ANOD receives information not only about its own internal classes but also about those types of connectors that expect partners in other organs. A node of class A is created with a PEND connector of type CONT-X; a node of class B is created with a complementary PEND connector of the same type. When both start SYGA — they find each other through the hierarchical bulletin board.

Thus, inter-organ connections are established through SYGA, not through SYGD. Each ANOD knows only its own classes and its own CONT types; coordination happens autonomously through the standard mechanism of synaptogenesis. This is consistent with biology: connections between different brain regions form predominantly through activity-dependent synaptogenesis, not through rigidly programmed projections.

5. 6. Life cycle of ANOD

The life cycle of one ANOD is a sequence of phases from creation to transition to functioning mode. All phases already described in Chapter 4 for morphogenesis as a whole are concretized here for one ANOD.

a) Creation of ANOD

ANOD is created at the splitting of MOVE on GATE. For each component of the COMP diagram, GATE initiates the creation of one ANOD: a UNON is allocated, the three sections are created (G, D=LOAI, M=LOMN), the MOVE fragment for that component is written to the M-section, the necessary D-components are loaded into the D-section (or, more precisely, d-relations to them in the ROOT repository are established), the G-section is initialized with the RULE matrix and initial AVEC quota.

b) Parsing

The first phase of ANOD's work is parsing its MOVE fragment. LOMN reads the fragment, validates references to D-components (do they exist in LOAI, are their signatures undamaged), computes the AVEC needs (how many nodes will need to be created, how many connections).

If resources are clearly insufficient for the morphogenesis of even a minimally viable variant of the organ, LOMN signals to GATE and requests a revision of quotas. GATE can redistribute AVEC between ANODs or report resource shortage up the hierarchy.

On successful parsing, LOMN forms the queue of morphogenetic tasks: the list of classes for instantiation, in the necessary order, with the necessary number of instances and necessary connections.

c) Active morphogenesis

From the task queue, LOMN launches the phases of morphogenesis described in Chapter 4: NRGN, SYGD, SYGA. These phases overlap in time. In parallel with them, verification works: LOMN periodically sends test signals through the already built part of the network, checks connectivity, registers the initial states of nodes in TRL.

The duration of active morphogenesis depends on the size of the organ. For a small organ (dozens of nodes), morphogenesis may complete in seconds. For a large organ (millions of nodes) — in hours or days. In both cases, the process is autonomous: LOMN does not need external control; it follows its MOVE fragment.

d) Maturation

As the main NRGN completes and most SYGD connections are established, ANOD enters the maturation phase. In this phase:

Most of the organ's nodes are already working — exchanging datagrams with neighbours, performing their functions, updating TRL.

SYGA continues in background mode: the remaining PEND connectors search for partners through the bulletin board, and some of them find them.

Activity-dependent correction works: connections between actively interacting nodes are strengthened; rarely used ones may be pruned.

LOMN gradually switches from active neurogenesis to coordination of work. Neurogenesis becomes a rare operation, occurring only when missing classes need to be built up.

e) Transition to functioning

When morphogenetic activity falls below a certain threshold, ANOD records the completion of morphogenesis and enters operational mode. This is the transformation OPNG → OPN described in Chapter 4. LOMN transforms from builder into administrator of the built organ.

From this moment, ANOD continues to work but in another capacity. Its main functions are: routing of datagrams between the nodes of its organ, management of operational AVEC quotas, security control, response to requests from other ANODs and from parent nodes.

The stem-cell function of LOMN is preserved in dormant mode. When necessary — damage, requirement of adaptation, release of an updated MOVE — ANOD can reactivate the morphogenetic functions and build up or rebuild part of the organ. This is a rare event but architecturally supported.

5. 7. Boundary between MOGE and GNET

The completion of morphogenesis of one organ — the transition of ANOD from OPNG into OPN — marks the boundary between the subject of MOGE and the subject of GNET. Before this boundary — morphogenesis: the process of formation of the network from a description. After — operation: the work of an already formed network.

a) What belongs to MOGE

The subject of MOGE is everything that happens between the placement of GERM on a platform and the completion of morphogenesis of all its organs:

Parsing of MOVE, splitting between ANODs, creation of ANODs, neurogenesis (NRGN), deterministic synaptogenesis (SYGD), autonomous synaptogenesis (SYGA), maturation, verification, transformation OPNG → OPN.

Distribution of D-components from ROOT to classes through d-relations. Establishment and distribution of AVEC along the tree ROOT → GATE → ANOD → subordinate nodes. Closure of the critical period and transition to operational mode.

Reactivation of morphogenetic functions when necessary — a rare event but still within MOGE by nature.

b) What belongs to GNET

The subject of GNET begins where the network is ready for operation:

Routing of datagrams between nodes at three levels (G1, G2, G3 — inside GATE, in the local network, across the internet, respectively). Protocols of establishment and maintenance of ordinary working relations (v-relations for values, b-relations for behaviour, and so on).

Management of operational assets — transfer of digital assets between nodes, registration of transactions in the distributed registry, notarization, smart contracts. This also includes the management of MOVEs themselves as assets in the registry (see Chapter 3).

Operational plasticity — correction of connections of already existing nodes based on accumulated experience. This is not the creation of new nodes nor the establishment of new deterministic connections, but the adaptation of an already working network. Operational plasticity relies on TRL and connector-usage statistics.

Security of the working network — immune response to illegitimate requests, cytokine signalling, distribution of CRL, revocation of compromised keys. Much of this uses mechanisms established at morphogenesis (the G-section of ANOD, AVEC₀ as immune marker) but functions actively in the operational phase.

c) Flexible boundary

The boundary between MOGE and GNET is functional, not temporal. In one and the same organism, some organs may already be in the operational phase (the subject of GNET) while others are still undergoing active morphogenesis (the subject of MOGE). This is a normal state of a large developing network.

The boundary can be crossed back: upon reactivation of morphogenetic functions for recovery or adaptation, an organ temporarily returns from the GNET mode to the MOGE mode. After the completion of reactivation, it goes back into operation.

This flexibility reflects the biological reality. In a developing organism, some systems are already working in their roles while others are still being formed. In an adult organism, most systems work, but some retain limited morphogenetic capabilities (neurogenesis in the hippocampus, renewal of epithelium, regeneration). Gativus inherits this structure.

5. 8. Conclusion

The architecture of morphogenesis in Gativus is organized as a hierarchy with a clear division of responsibility at each level. ROOT gives names and basic quotas, acting as the womb of the system. GATE receives GERM, supplements AVEC with hardware resources, and splits MOVE among ANODs. ANOD grows one functional organ, coordinating its entire morphogenesis autonomously.

Inside ANOD, the three-section architecture divides responsibility: the G-section controls, LOAI provides code, LOMN builds. This division excludes conflicts of functions and ensures architectural cleanliness. All three sections work inside one NDDI process without network overhead.

MOVE is split upon delegation from GATE to ANOD: each ANOD receives the fragment relating to its organ. The full MOVE remains with GATE and ensures the possibility of replacing damaged ANODs. Information about inter-organ connections is transferred through CONT types, and the connections themselves are established autonomously through SYGA.

The life cycle of ANOD passes through the phases of parsing, active morphogenesis, maturation, transition to functioning. The phases overlap in time, ensuring the parallelism of the process. The completion of morphogenesis — the transformation OPNG → OPN — marks the transition from the subject of MOGE to the subject of GNET.

The boundary between MOGE and GNET is functional, not temporal. One organism can contain organs in both modes simultaneously, and the transition between modes is reversible: reactivation of morphogenetic functions is possible when necessary.

On this, MOGE as a theoretical work concludes. The initial question — how a running network is obtained from a description — has received its answer through four chapters. Chapter 1 established GTR0 as the fourth transformation of Gativus, isomorphic to the other three. Chapter 2 introduced MOLD as the formal language of morphology description, and MOVE as a 16-dimensional vector. Chapter 3 resolved the question of the status of MOVE as an entity of dual nature — both component and asset. Chapter 4 described RTR0 as the morphogenesis procedure with its atomic units, phases, and parallelism. This chapter has completed the picture with the architecture in which this procedure is executed.

MOGE prepares the ground for GNET. Many terms and mechanisms used here with references to GNET — GORP, AVEC, CRL, GATN assets, the tree of registries — receive full specification in the next book. The boundary between MOGE and GNET fixed in section 5.7 delimits the subject of the two books: MOGE answers the question of the origin of the network, GNET — of its operation.