INFORMATION_SCHEMA was endorsed by the ISO/ANSI body long after the real RDBMS implementations moved into the market. As a result, some of the vendors implemented their own version of the metadata repository in a form of system tables.
The information from these tables can be gathered in a variety of ways, usually through views, provided by the vendor for just this purpose (the idea behind the INFORMATION_SCHEMA), through some vendor-supplied stored procedures or functions, or by making your best guess via direct querying of the underlying system tables.
Most vendors explicitly discourage users from accessing the system tables directly because their structure might change without any notice, and the information contained in the table is not guaranteed to mean what you think it should; in short, system tables are for the use of the system, views are for the users.
While introducing the standard is a step in the right direction toward uniform interface to metadata, it probably will not happen overnight (considering that INFORMATION_SCHEMA was first introduced in the SQL92 standard). The system catalog of the structure specified in the SQL92/99 standard is a requirement for achieving higher levels of the standard conformance; most of the RDBMS products are only Level 1 compliant and are likely to remain so in the nearest future.