Re: 9.4 release notes

Hi,

On 2014-05-04 08:46:07 -0400, Bruce Momjian wrote:
> Feedback expected and welcomed.  I expect to be modifying this until we
> release 9.4 final.  I have marked items where I need help with question
> marks.

Thanks for doing that work. Some comments inline:

+    <listitem>
+     <para>
+      Remove system column pg_class.reltoastidxid (Michael Paquier)
+     </para>
+
+     <para>
+      Instead use normal index access methods.
+     </para>
+    </listitem>

The explanation doesn't seem helpful to me. I'd just remove it as the
full explanation seems to be to complex and not actually interesting.


+   <sect3>
+    <title>Server</title>
+
+     <itemizedlist>
+
+      <listitem>
+       <para>
+        Have VACUUM properly report dead but not removable rows to the statistics collector (Hari Babu)
+       </para>
+
+       <para>
+        Previously these were reported as live rows.
+       </para>
+      </listitem>
+
+      <listitem>
+       <para>
+        Improve SSL renegotiation handling (Álvaro Herrera)
+       </para>
+      </listitem>
+
+      <listitem>
+       <para>
+        During immediate shutdown, send uncatchable termination signals to child processes that have not already
shutdown(MauMau,
 
+        Álvaro Herrera)
+       </para>
+
+       <para>
+        This reduces the likelihood of orphaned child processes after postmaster shutdown.
+       </para>
+      </listitem>
+
+      <listitem>
+       <para>
+        Improve randomness of the database system identifier (Tom Lane)
+       </para>
+      </listitem>
+
+      <listitem>
+       <para>
+        Allow background workers to be dynamically started and terminated (Robert Haas)
+       </para>

This is much more important than the previous features imo. Also, I'd
add the word "registered" in the above list.

+      <listitem>
+       <para>
+        Allow dynamic allocation of shared memory segments (Robert Haas, Amit Kapila)
+       </para>
+      </listitem>

Should also be earlier in the list.

+      <listitem>
+       <para>
+        Freeze tuples when tables are written with CLUSTER or VACUUM FULL (Robert Haas, Andres Freund)
+       </para>
+
+       <para>
+        This avoids later freezing overhead.
+       </para>
+      </listitem>

This sentence seems a bit awkward to me. Maybe "This avoids the need to
freeze tuples at some later point in time."

+      <listitem>
+       <para>
+        Improve speed of accessing many sequence values (David Rowley)
+       </para>
+      </listitem>

I think this description isn't accurate. Isn't it something like
"Improve speed of accesessing many different sequences in the same backend."

+      <listitem>
+       <para>
+        Use memory barriers to avoid some spinlock use (Heikki Linnakangas)
+       </para>
+      </listitem>

This is about 1a3d104475ce01326fc00601ed66ac4d658e37e5? If so I'd put it
as: "Reduce spinlock contention during WAL replay." or similar.

This imo deserves to be further up this list as this is one of the
biggest bottlenecks in HS node performance.

+      <listitem>
+       <para>
+        Add huge_pages configuration parameter to attempt to use huge translation look-aside buffer (TLB) pages on
Linux(Christian Kruse,
 
+        Richard Poole, Abhijit Menon-Sen)
+       </para>

I think this is too detailed - how about: "... to use huge pages ..."?

+       <para>
+        This can improve performance on large memory systems.
+       </para>
+      </listitem>

Should this be in the performance section?

+      <listitem>
+       <para>
+        Add configuration variable data_checksums to report whether data page checksums are enabled (Bernd Helmle)
+       </para>
+      </listitem>

Since this has been backpatched since it should probably be removed.

+      <listitem>
+       <para>
+        Use the last specified recovery_target if multiple are specified (Heikki Linnakangas)
+       </para>
+      </listitem>

This might want an entry in the compatibility section.

+      <listitem>
+       <para>
+        Add replication slots to report the WAL activity on streaming standbys (Andres Freund, Robert Haas)
+       </para>
+
+       <para>
+        Description?  Logical only?
+       </para>
+      </listitem>

No, it's not logical only. How about:
'Replication slots allow to reserve resources like WAL on the primary
that are needed by standby servers."

+      <listitem>
+       <para>
+        Improve return codes from external recovery commands (Peter Eisentraut)
+       </para>
+      </listitem>

This doesn't seem to be very descriptive. s/improve/report/?

+      <listitem>
+       <para>
+        Write WAL records of running transactions every 15 seconds ? (Andres Freund)
+       </para>
+      </listitem>

"Write WAL records about running transactions more frequently."

"This allows standby servers to startup faster and to cleanup resources
more aggressively."

+     </itemizedlist>
+
+     <sect4>
+      <title>Logical Change-Set Encoding</title>

s/Encoding/Extraction/

This probably deserves one entry about the new feature instead of
separate entries about individually commited parts.

REPLICA IDENTITY should probably documented separately. Other solutions
can use it as well.

+      <listitem>
+       <para>
+        Allow security barrier views automatically updateable (Dean Rasheed)
+       </para>

*to be?

+      <listitem>
+       <para>
+        Add functions pg_filenode_relation() and pg_relation_filepath() to do relation/relfilenode conversions (Andres
Freund)
+       </para>

pg_relation_filepath() isn't new. I think this should be something like:
"Add function pg_filenode_relation() to allow for more efficient
filenode to relation lookup."

+       <para>
+        These use a new pg_class index to speed lookups.
+       </para>
+      </listitem>

Don't think that's interesting for the release notes.

+       <listitem>
+        <para>
+         Have \d display disabled system triggers (Bruce Momjian)
+        </para>
+
+        <para>
+         Previously if you disabled all triggers, only user triggers would show as disabled.
+        </para>
+       </listitem>

Maybe mention that this could lead to broken foreign key checks after an
ALTER TABLE .. DISABLE TRIGGERS?.


+   <sect3>
+    <title>Source Code</title>
+
+     <itemizedlist>
+
+      <listitem>
+       <para>
+        Improve the way tuples are frozen, to preserve forensic information ((Robert Haas, Andres Freund)
+       </para>
+
+       <para>
+        Code that inspects tuple flag bits will need to be modified
+       </para>
+      </listitem>

Not sure if that's just a source code level change.

+      <listitem>
+       <para>
+        Remove SnapshotNow and HeapTupleSatisfiesNow (Robert Haas)
+       </para>
+
+       <para>
+        All existing uses have been switched to more appropriate snapshot types.
+       </para>
+      </listitem>
+
+      <listitem>
+       <para>
+        Use an MVCC snapshot (rather than SnapshotNow) for catalog scans (Robert Haas)
+       </para>
+      </listitem>

I wonder if those two shouldn't be one entry.

+      <listitem>
+       <para>
+        Use printf() modifier "z" to specify size_t values (Andres Freund)
+       </para>
+      </listitem>

s/Use/Add infrastructure to allow to use the %x printf modifier .../

+      <listitem>
+       <para>
+        Memory barrier changes?
+       </para>
+      </listitem>

Not sure what that refers to?

+      <listitem>
+       <para>
+        Remove spinlock support for unsupported platforms SINIX, Sun3, and NS32K (Robert Haas)
+       </para>
+      </listitem>

I'd put it as "Remove leftover code for the unsupported
... platforms.".

+      <listitem>
+       <para>
+        Rewrite duplicate_oids Unix hell script in Perl (Andrew Dunstan)
+       </para>
+      </listitem>

Heh. s/hell/shell/

+      <listitem>
+       <para>
+        Remove IRIX port (Robert Haas)
+       </para>
+      </listitem>

Should probably be next to the unsupported platforms list.

+      <listitem>
+       <para>
+        Have pg_stat_statements use a flat file for query text storage, allowing higher limits (Peter Geoghegan)
+       </para>
+
+       <para>
+        Also add the ability to retrieve all pg_stat_statements information except the query text.  This allows
programsto reuse the query
 
+        text already retrieved by referencing queryid.
+       </para>
+      </listitem>

This isn't an optional thing, is it?



Greetings,

Andres Freund

-- Andres Freund                       http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services