2 $Header: /cvsroot/pgsql/doc/src/sgml/history.sgml,v 1.8 2000/05/02 20:01:51 thomas Exp $
6 <title>A Short History of <productname>Postgres</productname></title>
9 The Object-Relational Database Management System now known as
10 <productname>PostgreSQL</productname> (and briefly called
11 <productname>Postgres95</productname>) is derived from the
12 <productname>Postgres</productname> package written at Berkeley.
14 development behind it, <productname>PostgreSQL</productname>
15 is the most advanced open-source database available anywhere,
16 offering multi-version concurrency control, supporting almost
17 all SQL constructs (including subselects, transactions, and
18 user-defined types and functions), and having a wide range of
19 language bindings available (including C, C++, Java, perl, tcl, and python).
23 <title>The Berkeley <productname>Postgres</productname> Project</title>
26 Implementation of the <productname>Postgres</productname>
27 <acronym>DBMS</acronym> began in 1986. The
28 initial concepts for the system were presented in
29 <xref endterm="STON86-full" linkend="STON86">
30 and the definition of the initial data model
32 <xref endterm="ROWE87-full" linkend="ROWE87">.
33 The design of the rule system at
34 that time was described in
35 <xref endterm="STON87a-full" linkend="STON87a">.
37 and architecture of the storage manager were detailed in
38 <xref endterm="STON87b-full" linkend="STON87b">.
42 <productname>Postgres</productname> has undergone several major releases since
43 then. The first "demoware" system became operational
44 in 1987 and was shown at the 1988 <acronym>ACM-SIGMOD</acronym>
45 Conference. We released Version 1, described in
46 <xref endterm="STON90a-full" linkend="STON90a">,
47 to a few external users in June 1989. In response to a
48 critique of the first rule system
49 (<xref endterm="STON89-full" linkend="STON89">),
52 (<xref endterm="STON90b-full" linkend="STON90b">)
54 released in June 1990 with the new rule system.
55 Version 3 appeared in 1991 and added support for multiple
56 storage managers, an improved query executor, and a
57 rewritten rewrite rule system. For the most part,
58 releases until <productname>Postgres95</productname> (see below)
59 focused on portability and reliability.
63 <productname>Postgres</productname> has been used
64 to implement many different
65 research and production applications. These include: a
66 financial data analysis system, a jet engine
67 performance monitoring package, an asteroid tracking
68 database, a medical information database, and several
69 geographic information systems.
70 <productname>Postgres</productname> has also been
71 used as an educational tool at several universities.
73 <ulink url="http://www.illustra.com/">Illustra Information Technologies</ulink>
75 <ulink url="http://www.informix.com/">Informix</ulink>)
77 the code and commercialized it.
78 <productname>Postgres</productname> became the primary data manager
80 <ulink url="http://www.sdsc.edu/0/Parts_Collabs/S2K/s2k_home.html">Sequoia 2000</ulink>
81 scientific computing project in late 1992.
85 The size of the external user community
86 nearly doubled during 1993. It became increasingly
87 obvious that maintenance of the prototype code and
88 support was taking up large amounts of time that should
89 have been devoted to database research. In an effort
90 to reduce this support burden, the project officially
91 ended with Version 4.2.
96 <title><productname>Postgres95</productname></title>
100 <ulink url="mailto:ayu@informix.com">Andrew Yu</ulink>
102 <ulink url="http://http.cs.berkeley.edu/~jolly/">Jolly Chen</ulink>
103 added a SQL language interpreter to <productname>Postgres</productname>.
104 <productname>Postgres95</productname> was subsequently released to
105 the Web to find its own way in the world as an
106 open-source descendant of the original <productname>Postgres</productname>
111 <productname>Postgres95</productname> code was completely
112 ANSI C and trimmed in size by 25%. Many
113 internal changes improved performance and maintainability.
114 <productname>Postgres95</productname> v1.0.x ran about 30-50%
115 faster on the Wisconsin Benchmark compared to
116 <productname>Postgres</productname> v4.2.
117 Apart from bug fixes, these were the major enhancements:
122 The query language <productname>Postquel</productname> was replaced with
123 <acronym>SQL</acronym> (implemented in the server).
124 Subqueries were not supported until
125 <productname>PostgreSQL</productname> (see below), but they
126 could be imitated in <productname>Postgres95</productname> with user-defined
127 <acronym>SQL</acronym> functions. Aggregates were
128 re-implemented. Support for the GROUP BY query clause was also added.
129 The <filename>libpq</filename> interface remained
130 available for <acronym>C</acronym>
137 In addition to the monitor program, a new program
138 (<application>psql</application>) was provided for interactive SQL queries
139 using <acronym>GNU</acronym> <filename>readline</filename>.
145 A new front-end library, <filename>libpgtcl</filename>,
146 supported <acronym>Tcl</acronym>-based clients. A sample shell,
147 pgtclsh, provided new Tcl commands to interface
148 <application>tcl</application>
149 programs with the <productname>Postgres95</productname> backend.
155 The large object interface was overhauled. The Inversion large objects were
156 the only mechanism for storing large objects.
157 (The Inversion file system was removed.)
163 The instance-level rule system was removed.
164 Rules were still available as rewrite rules.
170 A short tutorial introducing regular <acronym>SQL</acronym> features as
171 well as those of <productname>Postgres95</productname> was
172 distributed with the source code.
178 <acronym>GNU</acronym> make (instead of <acronym>BSD</acronym> make) was used
179 for the build. Also, <productname>Postgres95</productname> could be
180 compiled with an unpatched <productname>gcc</productname>
181 (data alignment of doubles was fixed).
189 <title><productname>PostgreSQL</productname></title>
192 By 1996, it became clear that the name "Postgres95" would
193 not stand the test of time. We chose a new name,
194 <productname>PostgreSQL</productname>, to reflect the relationship
195 between the original <productname>Postgres</productname> and the more
196 recent versions with <acronym>SQL</acronym> capability. At the same
197 time, we set the version numbering to start at 6.0, putting the
198 numbers back into the sequence originally begun by the
199 <productname>Postgres</productname> Project.
203 The emphasis during development of <productname>Postgres95</productname>
204 was on identifying and understanding existing problems in the backend code.
205 With <productname>PostgreSQL</productname>,
206 the emphasis has shifted to augmenting features and capabilities, although
207 work continues in all areas.
211 Major enhancements in <productname>PostgreSQL</productname> include:
217 Table-level locking has been replaced with multi-version concurrency control,
218 which allows readers to continue reading consistent data during writer activity
219 and enables hot backups from pg_dump while the database stays available for
226 Important backend features, including subselects, defaults,
227 constraints, and triggers, have been implemented.
233 Additional <acronym>SQL92</acronym>-compliant language features have been added,
234 including primary keys, quoted identifiers, literal string type coercion,
235 type casting, and binary and hexadecimal integer input.
241 Built-in types have been improved, including new wide-range date/time types
242 and additional geometric type support.
248 Overall backend code speed has been increased by approximately 20-40%,
249 and backend startup time has decreased 80% since v6.0 was released.
256 <!-- Keep this comment at the end of the file
261 sgml-minimize-attributes:nil
262 sgml-always-quote-attributes:t
265 sgml-parent-document:nil
266 sgml-default-dtd-file:"./reference.ced"
267 sgml-exposed-tags:nil
268 sgml-local-catalogs:("/usr/lib/sgml/catalog")
269 sgml-local-ecat-files:nil