A publisher is a database system that makes data available for replication.
A distributor is the database system that contains the distribution database,
or pseudodata, used to maintain and manage the replication.
A subscriber is a database system that receives replicated data and stores the replicated database.
* Publishers
The publisher consists of a Microsoft Windows system hosting a SQL Server database.
This database provides data to be replicated to other systems.
In addition, the SQL Server database keeps track of
which data has changed so that it can be effectively replicated.
The publisher also maintains information about which data is configured for replication.
Depending on the type of replication that is chosen,
the publisher does some work or little work during the replication process.
A replicated environment can contain multiple subscribers,
but any given set of data that is configured for replication, called an article, can have only one publisher.
Having only one publisher for a particular set of data does not mean
that the publisher is the only component that can modify the data
- the subscriber can also modify and even republish the data.
* Distributors
In addition to containing the distribution database,
servers acting as distributors store metadata, history data, and other information.
In many cases, the distributor is also responsible for distributing the replication data to subscribers.
The publisher and the distributor are not required to be on the same server.
In fact, you will likely use a dedicated server as a distributor.
Each publisher must be assigned a distributor when it is created,
and a publisher can have only one distributor.
NOTE
--------
Metadata is data about data.
Metadata is used in replication to keep track of the state of replication operations.
It is also the data that is propagated by the distributor to other members of the replication set
and includes information about the structure of data and the properties of data,
such as the type of data in a column (numeric, text, and so on) or the length of a column.
* Subscribers
Subscribers are the database servers that store the replicated data and receive updates.
Subscribers can also make updates and serve as publishers to other systems.
For a subscriber to receive replicated data, it must subscribe to that data.
Subscribing to replication involves configuring the subscriber to receive that data.
A subscription is the database information to which you are subscribing.
Source: Microsoft SQL Server 2000 Administrator's Companion Book