XML Support - The relational database engine can return 
data as Extensible Markup Language (XML) documents. 
Additionally, XML can also be used to insert, update, and 
delete values in the database. (for xml raw - to retrieve 
output as xml type) 
User-Defined Functions - The programmability of Transact-
SQL can be extended by creating your own Transact-SQL 
functions. A user-defined function can return either a 
scalar value or a table.
 
Indexed Views - Indexed views can significantly improve the 
performance of an application where queries frequently 
perform certain joins or aggregations. An indexed view 
allows indexes to be created on views, where the result set 
of the view is stored and indexed in the database. 
New Data Types - SQL Server 2000 introduces three new data 
types. bigint is an 8-byte integer type. sql_variant is a 
type that allows the storage of data values of different 
data types. table is a type that allows applications to 
store results temporarily for later use. It is supported 
for variables, and as the return type for user-defined 
functions. 

INSTEAD OF and AFTER Triggers - INSTEAD OF triggers are 
executed instead of the triggering action (for example, 
INSERT, UPDATE, DELETE). They can also be defined on views, 
in which case they greatly extend the types of updates a 
view can support. AFTER triggers fire after the triggering 
action. SQL Server 2000 introduces the ability to specify 
which AFTER triggers fire first and last. 

Multiple Instances of SQL Server - SQL Server 2000 supports 
running multiple instances of the relational database 
engine on the same computer. Each computer can run one 
instance of the relational database engine from SQL Server 
version 6.5 or 7.0, along with one or more instances of the 
database engine from SQL Server 2000. Each instance has its 
own set of system and user databases. 

Index Enhancements - You can now create indexes on computed 
columns. You can specify whether indexes are built in 
ascending or descending order, and if the database engine 
should use parallel scanning and sorting during index 
creation.