The operation that can be invoked on a unit/module/class/etc that executes its principle objective and responsibility.


Applicative is similar to "application", as in "application logic". Because services are not applications, the term "application logic" is out of place when talking about service implementation. The term "applicative", as in "applicative logic" is more appropriately used in this case.


Autonomy is arguably the most critical quality of service design. It's the thing that can dictate whether you end up with a service architecture, or a web of web services organized as a distributed monolith. Autonomy is the quality that allows services to remain online when services that they communicate with become unavailable. Autonomy has the most profound influence on the design of services, the things that are inside of a service, and the things that are outside of a service. Without autonomy, services arguably aren't services at all.

Cardinal ID

The cardinal ID is the first ID that appears in an entity stream's compound ID. If the stream name has only a single ID, then the cardinal ID is that single ID.


A category is a group of streams all related to the same type of entity. An analogy would be a class in an object-oriented system.

Category Stream

A stream that contains the messages for every individual, identified stream in the category. For example, given the streams account-123, and account-456, the category stream, account, contains messages from both individual streams in the account category.

Category Type

A category type is a qualifier added to a category name that is used to differentiate between categories used for various specific purposes.


A command is a message that represents an instruction sent to a component. A command is an input message. Commands are sent directly to components. They are a unidirectional transmission from one endpoint directly to another.

Command Stream

A stream that contains only command messages. Command streams are effectively the input queues that services read their commands from.


A package of code that represents a single business concern and/or business process, such as an account, a sale, and so forth. It's all the code wrapped around that concern to operationalize it. It has one or more entities, but very few. A component is often packaged as a RubyGem. A component is hosted either by itself or with other components in a service.

Component Host

The component host is the outermost surface of an Eventide service. It runs right next to the operating system. It is the physical "service". It loads components into itself. It coordinates the threads that the consumers run on. It provides graceful shutdown for the components hosted in a service.

Compound ID

A compound ID is the ID part of an entity stream name that is composed of multiple values.

Compound Category Type

A compound category type is a category type that is composed of multiple values to indicate the category's multiple purposes.


The consumer is the infrastructure that runs a subscription reader over streams, and plugs-in the handlers to those subscriptions to receive the messages they subscribe to.


An endpoint is the source process that a message comes from or the destination process where it is sent. The term endpoint is often used informally as a substitute for service.


The entity is the core logic of the component. It's the domain model. It's analogous to a Rails model, but that's only an analogy. One entity's data is contained in a single, individual event stream. An entity collects any important information from the events in the entity's stream. It's also a data object. An entity has a corresponding projection, which is used to copy data from the events to the entity object, either by setting the entity's attributes or by invoking its methods. The data that an entity has collected is also used in handler code to determine whether a command should be processed.

Entity Cache

When entities are retrieved via the entity store, they are cached in-memory, and, optionally on disk.

Entity Projection

An entity's data comes from events. An event is a thing that has happened in a process, or it's a change that has been made to an entity's state. Often, these two are one and the same. The projection is a mechanism that receives events as inputs, and modifies the state of an entity. When an entity is "retrieved", its events are applied to it by the entity projection.

Entity Snapshot

An entity snapshot is saved to disk (typically, in the message store) by the entity store as a performance optimization for warm-up time of the entity cache. The store saves a snapshot based on an assignable interval of events. For example, if the interval is set to 100, the store will save an entity's snapshot every 100 events that the store reads and projects onto the entity.

Entity Stream

A stream that contains the messages for a single, identified entity. For example, a single account with the ID 123 has all of its events in its own stream, and that stream contains only the events for account 123.

Entity Store

The entity store is used to do the retrieval of entities. It coordinates a reader, an entity projection, and an entity cache. In reads only new events for an entity, projects them on to the cached entity, and updates the cache with the result of the projection. It may optionally record cached snapshots of entities to disk once every n events read and projected.


A message that represents the results and effects of some command having been processed. It's a record of something that has happened or that has been processed. An entity's state is gathered from the events that have been written about the entity. Each activity of an entity's life is recorded as an event, and written to the entity's stream.

Event Sourcing

Event sourcing is a style of entity storage that relies on reading entity data from a series of events that are recorded in a stream that holds the events for a single entity. An entity's data is stored by writing events to its stream. When the entity is "retrieved" from the entity store, the events are passed through an entity projection, which populates the entity's attributes.

Event Stream

A stream that contains only event messages. Each entity instance maps to a single, individual event stream. An entity's data is collected from the events in its stream.

EventStore (Database)

One of the databases supported by Eventide. Event Store is an open source, special-purpose database designed specifically for high-availability event storage and event processing.

Event Store (Pattern)

A database specifically structured for the storage and retrieval of events. See also: message store.


Handlers have the business logic for processing the messages that are inputs to a component's consumers. Handlers are loosely analogous to controllers in a web MVC app.


A message is a data structure that represents either an instruction to be passed to a process (command), or a record of something that has happened (event) - typically in response to the processing of a command. Messages are written to the message store, and then read by consumers that then dispatch messages to handlers.

Message Queue and Message Bus

A form of message transport that relies on more elaborate mechanisms and protocols for moving messages through a system and keeping track of which messages have already been processed. These technologies are notably failure-prone and can cause subtle errors where messages are not delivered and the delivery failure can go undetected by operators, developers, and testers.

Message Reader

Retrieves ordered batches of messages from a single stream, processing each message, one-by-one, into a block that can be passed to the reader's actuator.

Message Store

Unlike a message broker, a message store acts not only as a message transport, but also a durable store of messages that have been written. A service receives commands by reading new messages that have been written to the message store. Events are written to the message store in response to handling the inbound commands. Entity state is gathered from the event data that is written to the message store's message streams.

Message Transport

Any technology that is used to send (transport) messages from one process to another. For example: a queue, a broker, or a raw network transport protocol like TCP or HTTP.

Message Writer

Writes a message, or a batch of messages, to a single stream.


Provenance is the chronology of the custody or location of a historical object. It's used in the context of messaging workflows to describe metadata of preceding messages in a sequence of messages in a causal chain of messages.


Pub/Sub is a messaging pattern where publishers produce events and subscribers consumer them. In Pub/Sub, events are written without the publisher's knowledge of the subscribers.


In its strictest sense, a service is an autonomous unit of system operations that is made up of components that the service hosts.


A stream is the fundamental unit of storage that a message is stored in. When a message is written to the store, the writer must also be given the stream name that the event will be stored in. A stream is only stored in a single message store.

Stream Name

A stream's name not only identifies the stream, but also its purpose. A stream name is a string that optionally includes an ID that is prefixed by a dash (-) character. The part of the stream preceding the dash is the category, and the part following the dash is the ID. A stream with an ID is typically an event stream containing the events for the entity with that ID. For example: account-123 is a stream for the account entity with an ID of 123. The account stream name is the category stream name that contains all of the aggregated messages from all individual, identified account streams. See the overview of stream names in the user guide for more information.

Last Updated:
Contributors: Scott Bellware, Danil, Matt Quinn, Will Howard