This term relates to the process of identifying the significant events in the transaction life-cycle, including error-processing, auditing, save-points, nested transactions, and distribution paths for distributed transactions.
I think you may find more results with the synonymous term Transaction Life-cycle Analysis.
Check out DML - Data Manipulation Language. This includes CRUD operations (Create (i.e. INSERT - not to be confused with the CREATE TABLE statement below), Read (i.e. SELECT), Update, Delete) on Tables. It's what people associate most classically with SQL.
Compare this with DDL - Data Definition Language. DDL involves the CREATE TABLE and ALTER TABLE operations (among others - related to INDEXes amongst other things).
For completeness, there is also DCL - Data Control Language. DCL is for the control of access to tables - GRANT and REVOKE (i.e. GRANT READ ON schema_name.table_name TO user_name.
The SQL standard is an attempt to homogenise DML among various vendors - its success has been limited so far. Many servers implement non-standard extensions, their own Stored Procedure language, DDL and DCL. Take a look at the "Procedural extensions" section on the SQL page above.
Changes to a schema (from the same version of the same server to another or even within the same instance) are simply called DDL - CREATE TABLE, ALTER TABLE, DROP TABLE.
Changes to a schema brought about (normally) by an upgrade (e.g. Oracle 9 to Oracle 11) are called migrations - going the other way is called either downgrading or backward migration.
Changes to a schema due to a change of server (i.e. Oracle to SQL Server) are called ports ("porting"). These are normally quite difficult and involved due to the differences between servers mentioned above.
Best Answer
Depends on who you ask.
Is it in terms of new features, etc.? For example, a 32-bit RDBMS system could be considered obsolete nowadays, because 64-bit is common, and most engines run safely on a 64-bit platform, with access to more memory.
Is it in terms of scaling? For example, an old design that doesn't scale well, might be considered obsolete.
A database with queries that use old-style joins, could be considered obsolete, and in this case would prevent upgrades to newer technology without changing the code.
Some wonks like to say that SQL itself is obsolete, that NoSQL technology is better suited to certain problem areas like unstructured data. I personally wouldn't say this is the right use of "obsolete", but that could be an example too.