PostgreSQL Weekly News - June 18th 2003

== PostgreSQL Weekly News - June 18th 2003 ==

After last week's mention of running PostgreSQL on an Opteron with
Debian Linux in 64-bit node, I received a couple of emails about other
people using PostgreSQL on 64-bit hardware. One in particular was a note
that SuSE Linux Enterprise Server 8 for AMD64 has been shipping
with support for PostgreSQL 7.2.4 since April, and is actively being
maintained.

With that said, we are now less than two weeks away from feature freeze
(July 1st). Bruce Momjian has gone through a number of patches, if you
plan to add something new for the next release it needs to be submitted
as soon as possible.

Speaking of the next version, we have several updates on the win32
front. Jan Wieck is still working on the exec() bit, and we've started
to see several folks working through compile issues on -hackers. There
are promising signs, like reports of successful builds of various
flavors, but there is still a fair bit needed like signal handling, so
for now we're keeping our fingers crossed that this will make it in.

Ecpg saw a number of improvements this week, including fixes to make
sure a variable is no longer referenced when it is removed, fix a
counting bug in parsing "->" operator, and remove an accidentally
committed debugging function. Some Informix compatibility tweaks were
also done, a change was made to the Makefile to ensure all header files
are installed, some version numbering issues were worked out, Fetch
without INTO was enabled, ecpg thread safty should now be worked out,
some cleanups in for prototypes were done, and the ECPG Todo list was
updated.

The new pg_autovacuum features has been steadily improving, recent
changes include fixing solaris compile and crash issues, decoupling
vacuum analyze and analyze thresholds, detach from tty, improved logging
layout, a more conservative default configuration, improved, expanded
and updated README, and a few other bug fixes as well. The bulk of this
work is thanks to Mattew T. O'Conner and Christopher Browne. Matt had
this to say about the current state of things "At this point I think I
have brought pg_autovacuum and its client side design as far as I think
it should go. It works, keeping file sizes in check, helps performance
and give the administrator a fair amount flexibility in configuring it.
Next up is to do the FSM based design that is integrated into the
back end."

We've also reaped some early fruits from Manfred Koizars work on nested
transactions. He has changed page buffer pooling and the
"least-recently-used" strategy from clog.c to slru.c. Along with this
SlruRecentlyUsed, formerly ClogRecentlyUsed, was changed to skip
incrementing lru_counts, if slotno is already the LRU slot, thus saving
a few CPU cycles. SimpleLru will be used by pg_subtrans (part of the
nested transactions project), so the main purpose of this patch is to
avoid future code duplication. SerializableSnapshot and QuerySnapshot
are also no longer malloc'ed and free'd for every transaction or
statement, instead they these data structures into static memory, thus
saving a few CPU cycles and two malloc calls per transaction (or in
isolation level READ COMMITTED per query).

Sean Chittenden has been doing some recent work on OpenSSL, and as a
result found several ways to bolster PostgreSQL's SSL support.
PostgreSQL now sets the cipher list that it allows, it's renegotiation
code was improved to be "text book correct", and the rate of
renegotiation was set to a more reasonable level. He mentioned that
should he run across anything else in his current work, patches would
follow; thanks Sean!

Two recent changes to GROUP BY were put in. The first allows the GROUP
BY clause to adopt ordering operators from ORDER BY when both clauses
specify the same targets, rather than always using the default ordering
operator. This allows 'GROUP BY foo ORDER BY foo DESC' to be done with
only one sort step. The other change allows GROUP BY, ORDER BY, and
DISTINCT targets to be unknown literals, silently resolving them to type
TEXT. This is comparable to what we do when faced with UNKNOWN in CASE,
UNION, and other contexts and fixes an ordering operator error.

Several of the admin tools saw changes this week. Some inconsistencies
in pg_dump and pg_restore's handling of SQL identifiers was fixed.
Pg_dump was also fixed so that the combination of --schema and --table
will dump exactly one table from exactly one schema. Ian Barwick added
tab completion of \encoding, and Greg Sabino Mullane enhanced psql's
HTML mode out put. The output now validates as HTML 4.01 Strict, XHTML
1.0 strict, and XHTML 1.1 (assuming you wrap it in a valid html/body
document). Clusterdb and vacuumdb were also converted into C programs.
Along similar lines the linux start script had restart added and now
also has the "-l $PGLOG" option.

Theres been a bit of discussion (and some fixes) regarding python
support as of late. The fixes include improving the way which python
input converter to use is decided, and fixing some issues in setup.py.
The discussion revolves around the fact that there have been a number of
changes in Python versions greater than 2.2.3 which makes our current
support of plpython somewhat broken. Some folks have talked about
revamping the language to be used only as an untrusted language, if your
interested in helping to maintain this package please post a note to the
pgsql-interfaces list. This is something that the core team is not
likely to work on so it would be really great if someone can step up to
the plate.

Intarray has also been spruced up; changes include a bugfix for
int[]-int[] operation, split _int.c to several files (_int.c now is
unused), use special type for index storage for intarray in
gist__intbig_ops opclass, and several several missing intarray files
were added. On a rather unrelated note, there also been a bit of
sprucing up in postgresql libraries thread support. Changes here
involved adding thread-enable compile variables into libpq, adding
--with-threads configure option to control threaded libpq, adding
thread.c for libpq threading (and hooking it into libpq/configure),
adding the ability to handle threading in two more gethostbyname calls,
and moving thread os defines into template files.

Even with the number of changes already listed, we still have a sizable
miscellaneous fixes list this week. Andrew Dunstan and Kurt Roeckx have
done some work to allow for CIDR netmasks in pg_hba.conf. It allows two
address/mask forms; address/maskbits or address netmask (as now).
They've also done a bit of IPv6 clean up as well. Chris Campbell has
added Rendezvous support for the postmaster, information schema now
represents grant options, and the check for newly created matching PK
rows for on update action was tweak to move it to after checking for
NULL in the old row and comparing the values. Ben Lamb worked up a patch
to greatly increase the speed of libpq's PGunescapeBytea function, based
on some work by Billy Allie. The SQL function executor was fixed for
cases where last command of a function is not a SELECT (Which use to not
be allowed, but is now is. Several cryptic 'Unknown kind of return type'
messages were replaced with some hopefully a more useful messages. Some
bugs in interval-to-time conversion were fixed; the HAVE_INT64_TIMESTAMP
case did not work at all, and neither case behaved sanely for negative
intervals. Some error line numbers reported for errors in
plpgsql_parse_word and its siblings were improved as well. FLOAT(p) now
measures the precision p in bits, not decimal digits, to match the SQL
standard and finally two new GUC parameters were added:
log_min_duration_statement (which is SUSET) and add add_missing_from.
The add missing from gets special mention for it's submitter, Nigel J.
Andrews, since it was an official "TODO" item. ;-)

We've also had various documentation fixes this week, including
documenting why certain GUC variables aren't in the postgresql.conf,
documentation of the -h client flag being usable for a socket directory
as well as a host name, an update to the anonymous CVS instructions,
some documentation of the default Linux autocommit behavior (with
workaround), and a mention of the need for swap on Linux, and kill -9.
It might just be me but I think 7.4 docs are really shaping up quite
nicely.

== PostgreSQL Product News ==

PeerDirect programme enables distributed retail solution for multiple
stores
http://www.itweb.co.za/sections/business/2003/0306170811.asp

Mammoth PostgreSQL 7.3.2 Rekall Combo Pack Announced
http://www.businesswire.com/cgi-bin/cb_headline.cgi?&story_file=bw.061703/231685792

Synchronicity Enhances Scalability, Performance of Developer Suite
http://www.businesswire.com/cgi-bin/cb_headline.cgi?&story_file=bw.061603/231675828

== PostgreSQL In the News ==

Are Open Source Databases Following in Linux' Footsteps?
http://boston.internet.com/news/article.php/2221901

Review: Pogo Linux StorageWare S212 Server
http://newsforge.com/article.pl?sid=03/06/13/183237&mode=thread&tid=7

== Upcoming Events ==

MIT Open Source Conference: Cambridge, Massachusetts: June 19-20
Bruce Momjian will be involved in a panel discussion June 19th
http://opensource.mit.edu/conference.html

Open Source Conference: Portland, Oregon: July 7-11
A PostgreSQL track is available this year
http://conferences.oreillynet.com/os2003

LinuxTag: Karlsruhe Convention Center, Germany: July 10-12
A PostgreSQL Talk will be given on July 12th
http://www.linuxtag.org/2003/en/index.html

== PostgreSQL Weekly News - June 18th 2003 ==
Don't forget to read Elein Mustain's Weekly Summary of the PostgreSQL
General Mailing List http://www.varlena.com/GeneralBits/

On the Web:
http://www.postgresql.org
http://advocacy.postgresql.org