pgsql-server: Update that 8.0 will support MS Win natively.

Bruce Momjian wrote:

Update that 8.0 will support MS Win natively.

I've noticed that when we mention Win32 native support in 8.0 (both here
and in the release notes), we don't say it is "considered to be beta",
or "less well-tested than PostgreSQL for Unix systems", or some other
caveat. While I haven't used the Win32 port myself, I still think such a
disclaimer would be wise: the Win32 port has received a tiny fraction of
the testing that the Unix port has. So we really *don't* know how
unstable/buggy it may be, and until we have more data, I think
under-promising is a good idea.

(... or has this already been discussed over the summer?)

-Neil

Neil Conway wrote:

Bruce Momjian wrote:

Update that 8.0 will support MS Win natively.

I've noticed that when we mention Win32 native support in 8.0 (both here
and in the release notes), we don't say it is "considered to be beta",
or "less well-tested than PostgreSQL for Unix systems", or some other
caveat. While I haven't used the Win32 port myself, I still think such a
disclaimer would be wise: the Win32 port has received a tiny fraction of
the testing that the Unix port has. So we really *don't* know how
unstable/buggy it may be, and until we have more data, I think
under-promising is a good idea.

(... or has this already been discussed over the summer?)

I think we need to actually find someone who reports a problem before
stating something. I don't see how we can assume it is unstable without
some feedback.

Everyone knows it is is a new feature so I assume they will realize
their might be some rough edges.

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

Bruce Momjian <pgman@candle.pha.pa.us> writes:

I think we need to actually find someone who reports a problem before
stating something. I don't see how we can assume it is unstable without
some feedback.

You have *got* to be kidding. It's a new port with major changes
needed, and you're going to assume it is bulletproof? We do not work
that way on this project --- we err on the side of conservatism not
optimism.

In any case, it's not like we haven't been seeing plenty of bug reports
from users of the Windows beta. Many of these may be pilot error, but
I wouldn't care to assume they all are.

regards, tom lane

Bruce Momjian wrote:

I think we need to actually find someone who reports a problem before
stating something. I don't see how we can assume it is unstable without
some feedback.

On the contrary, I don't see how we can assume it IS stable without any
evidence (which is effectively what we're doing if we release 8.0.0
without any special note about how stable we expect Win32 to be: we're
treating unix and win32 equally, when they clearly are not from the POV
of testing and maturity).

This is enterprise software -- I think it would be wise for us to be
conservative about what we promise our users.

Everyone knows it is is a new feature so I assume they will realize
their might be some rough edges.

If "everyone knows", why not document it?

-Neil

Neil Conway wrote:

Bruce Momjian wrote:

I think we need to actually find someone who reports a problem before
stating something. I don't see how we can assume it is unstable without
some feedback.

On the contrary, I don't see how we can assume it IS stable without any
evidence (which is effectively what we're doing if we release 8.0.0
without any special note about how stable we expect Win32 to be: we're
treating unix and win32 equally, when they clearly are not from the POV
of testing and maturity).

This is enterprise software -- I think it would be wise for us to be
conservative about what we promise our users.

What makes it more different from saying PITR, NT, or tablespaces might
have bugs because those are new features too. What is the distinction?

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

On Mon, 23 Aug 2004, Bruce Momjian wrote:

Neil Conway wrote:

Bruce Momjian wrote:

I think we need to actually find someone who reports a problem before
stating something. I don't see how we can assume it is unstable without
some feedback.

On the contrary, I don't see how we can assume it IS stable without any
evidence (which is effectively what we're doing if we release 8.0.0
without any special note about how stable we expect Win32 to be: we're
treating unix and win32 equally, when they clearly are not from the POV
of testing and maturity).

This is enterprise software -- I think it would be wise for us to be
conservative about what we promise our users.

What makes it more different from saying PITR, NT, or tablespaces might
have bugs because those are new features too. What is the distinction?

they are new features, not new ports ...

Win32 is a new platform that we are supporting, and the likelihood of
someone finding a bug somewhere in the tens of thousands of lines of code
that is "windows specific" right now is fairly high ...

adding NT/PITR adds a feature that unless someone *really* screwed up,
doesn't have the potential of finding a bug *anywhere* in our code other
then where they tie into it, so its alot less of an impact overall ...

And that probably isn't worded as well as I'd like ... basically, *none*
of our code is *well tested* on Windows, where is *most* of our code is
well tested under Unix ... the only common "not well tested" code is the
new features we add during the release ...

----
Marc G. Fournier Hub.Org Networking Services (http://www.hub.org)
Email: scrappy@hub.org Yahoo!: yscrappy ICQ: 7615664

On Tue, Aug 24, 2004 at 08:50:44AM -0300, Marc G. Fournier wrote:

adding NT/PITR adds a feature that unless someone *really* screwed up,
doesn't have the potential of finding a bug *anywhere* in our code other
then where they tie into it, so its alot less of an impact overall ...

At least NT is so invasive that it has to potential to really screw
up ... in fact several ugly bugs have been detected and corrected already
(thanks guys!), not sure how many more are still waiting to be found.

--
Alvaro Herrera (<alvherre[a]dcc.uchile.cl>)
"La principal caracter�stica humana es la tonter�a"
(Augusto Monterroso)

Bruce Momjian wrote:

I think we need to actually find someone who reports a problem before
stating something. I don't see how we can assume it is unstable without
some feedback.

My count is three votes in favour of adding a disclaimer (myself, Tom,
Marc -- possibly Alvaro) and only one vote against (Bruce). Does anyone
think we ought to ask for opinions on -hackers? If not, I'll submit a
doc patch adding the disclaimer to the release notes.

Cheers,

Neil

Neil Conway wrote:

Bruce Momjian wrote:

I think we need to actually find someone who reports a problem before
stating something. I don't see how we can assume it is unstable without
some feedback.

My count is three votes in favour of adding a disclaimer (myself, Tom,
Marc -- possibly Alvaro) and only one vote against (Bruce). Does anyone
think we ought to ask for opinions on -hackers? If not, I'll submit a
doc patch adding the disclaimer to the release notes.

Is the release notes even the right place for it? When you read the
release notes after 8.1 is released, do you want to read that Win32
had possible problems in 8.0? Perhaps we should mention it in the
release announcement instead?

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

Bruce Momjian <pgman@candle.pha.pa.us> writes:

Is the release notes even the right place for it? When you read the
release notes after 8.1 is released, do you want to read that Win32
had possible problems in 8.0?

Sure you do. The release notes are historical material. There is
plenty of stuff in there that's totally irrelevant now, but we have
not (and I trust will not) gone back and removed entries that are no
longer significant.

Perhaps we should mention it in the
release announcement instead?

Why are you so insistent on sweeping this point under the rug? I'd much
rather underpromise and overdeliver than the reverse. We have plenty of
reason to be suspicious of the native Windows port at this stage ...
if you think it's going to be problem-free, I refer you to
http://archives.postgresql.org/pgsql-bugs/2004-08/msg00307.php
for the counterexample du jour. You'd be nuts to think that we'll find
every one of these issues before 8.0 release.

regards, tom lane

Tom Lane wrote:

Bruce Momjian <pgman@candle.pha.pa.us> writes:

Is the release notes even the right place for it? When you read the
release notes after 8.1 is released, do you want to read that Win32
had possible problems in 8.0?

Sure you do. The release notes are historical material. There is
plenty of stuff in there that's totally irrelevant now, but we have
not (and I trust will not) gone back and removed entries that are no
longer significant.

Perhaps we should mention it in the
release announcement instead?

Why are you so insistent on sweeping this point under the rug? I'd much
rather underpromise and overdeliver than the reverse. We have plenty of
reason to be suspicious of the native Windows port at this stage ...
if you think it's going to be problem-free, I refer you to
http://archives.postgresql.org/pgsql-bugs/2004-08/msg00307.php
for the counterexample du jour. You'd be nuts to think that we'll find
every one of these issues before 8.0 release.

Agreed, but we just found NT problems today too. The only argument I
can accept is that somehow platform bugs are harder to discover than our
other code bugs.

My concern is that I would like to have some actual facts before making
statements. I am not good on going on guesses, I guess. :-)

The open Win32 bugs have been there for quite some time, though the one
you quote is quite new so maybe platforms bugs are harder to find.

This is the same issue I had with removing NT4 as supported. I needed
to hear facts about what didn't work before making a statement because
when I go on guesses, I am usually wrong and have to somehow take it
back, and in this project, it is hard to retract things effectively.

As an example of the NT4 issue, it turns out it works just fine, just
that it doesn't support tablespaces, and the installer doesn't work.
Once I had that info I could adjust the release notes to just not
mention it in the list of platforms.

What text are people suggesting? "This is a new port and might have
bugs that will be fixed in minor releases?" Or, "This port is
experimental and you would be crazy to use it in production"? :-)

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

Bruce Momjian <pgman@candle.pha.pa.us> writes:

Agreed, but we just found NT problems today too. The only argument I
can accept is that somehow platform bugs are harder to discover than our
other code bugs.

The point isn't that they're harder to discover or not, it's that they
are new. 8.0 as a whole will have plenty of bugs that we will not find
till after release :-( (not happy about that statement, but it's a fact).
But we can be pretty certain that most of the bugs affecting Unix
platforms will be generic cross-platform bugs that also fail on the
Windows port. On *top* of those issues, the Windows port will have its
own problems. It's the merest wishful thinking to suppose that the
Windows port will be as stable as the longer-established ports. I do
not say that it won't or can't get there ... but I do say that it won't
happen on day zero, and we ought to be forthright about admitting that.

My concern is that I would like to have some actual facts before making
statements. I am not good on going on guesses, I guess. :-)

Wake up and smell the flowers, Bruce. There is plenty of evidence for
my position already in the pgsql-bugs archives, including the latest
item that I pointed you to just up-thread. You are being deliberately
clueless.

What text are people suggesting? "This is a new port and might have
bugs that will be fixed in minor releases?" Or, "This port is
experimental and you would be crazy to use it in production"? :-)

I'd go for (b) ;-) ... but then again I think anyone would be crazy to
use Windows for production anytime ;-). But whatever your opinion on
that, it would be irresponsible not to point out that this is a new port
that is certain to have more than its share of problems.

regards, tom lane

Bruce Momjian wrote:

What text are people suggesting?

How do people feel about the attached doc patch?

Cheers,

Neil

I have shortened your paragraph to:

Because Win32 is significantly different from the Unix platforms
supported in previous releases, this port might have more bugs
than other supported platforms in this release. Please
test it thoroughly before using it in production.

and made the change in release.sgml. Do we need more?

Other wording I considered was:

"we expect this port not to be as bug-free as other supported platforms"

but that is close to a double-negative.

In adding this now, I would also like to see a notice in a future
release at the point we think Win32 is as bug-free as Unix (or as
bug-free as the platform allows). (And this might happen in a minor
8.0.X release.)

Again, my concern about this stuff is that you later have to undo what
you said.

---------------------------------------------------------------------------

Neil Conway wrote:

Bruce Momjian wrote:

What text are people suggesting?

How do people feel about the attached doc patch?

Cheers,

Neil

Index: doc/src/sgml/release.sgml
===================================================================
RCS file: /Users/neilc/local/cvs/pgsql-server/doc/src/sgml/release.sgml,v
retrieving revision 1.288
diff -c -r1.288 release.sgml
*** doc/src/sgml/release.sgml	24 Aug 2004 00:06:50 -0000	1.288
--- doc/src/sgml/release.sgml	26 Aug 2004 05:43:32 -0000
***************
*** 28,47 ****
<listitem>
<para>
This is the first <productname>PostgreSQL</productname>
!         release to natively run on Microsoft Windows as a server. It
!         can run as a Windows service. This release supports NT-based
!         Windows releases like Win2000, XP, Win2003.  Older releases
!         like Windows 95, 98, and ME are not supported because these
!         operating systems do not have the infrastructure to support
!         <productname>PostgreSQL</productname>.  A separate installer
!         project has been created to ease installation on Windows:
<ulink url="http://pgfoundry.org/projects/pginstaller">
http://pgfoundry.org/projects/pginstaller</ulink>.
</para>
!        <para>
!         Previous releases required the Unix emulation toolkit Cygwin for
!         Win32 server support. <productname>PostgreSQL</productname>
!         has always supported clients on Win32.
</para>
</listitem>
</varlistentry>
--- 28,65 ----
<listitem>
<para>
This is the first <productname>PostgreSQL</productname>
!         release to natively run on Microsoft Windows as a server. This
!         release supports Windows 2000, XP, and 2003. Older releases of
!         Windows are not supported because they lack the infrastructure
!         necessary to support <productname>PostgreSQL</productname>. A
!         separate installer project has been created to allow for easy
!         installation on Windows:
<ulink url="http://pgfoundry.org/projects/pginstaller">
http://pgfoundry.org/projects/pginstaller</ulink>.
</para>
! 
!        <warning>
!         <para>
!          This is the first release of <productname>PostgreSQL</> to
!          natively support Win32. As a result, the Windows port of
!          <productname>PostgreSQL</> is less mature and likely less
!          stable than <productname>PostgreSQL</> on Unix. While we have
!          done our best to produce a high-quality release for Windows,
!          users accustomed to the stability and maturity of
!          <productname>PostgreSQL</> on Unix may not be satisfied by
!          the current state of <productname>PostgreSQL</> on
!          Windows. We expect the maturity of the Windows port to
!          significantly increase in the future. We encourage all
!          prospective users of <productname>PostgreSQL</> on Windows to
!          carefully evaluate the database system before putting it into
!          production use.
!         </para>
!        </warning>
! 
!        <para>
!         Previous releases required the Unix emulation toolkit Cygwin
!         for Win32 server support. <productname>PostgreSQL</> has
!         supported clients on Win32 for a long time.
</para>
</listitem>
</varlistentry>

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

Bruce Momjian wrote:

I have shortened your paragraph to:

Because Win32 is significantly different from the Unix platforms
supported in previous releases, this port might have more bugs
than other supported platforms in this release. Please
test it thoroughly before using it in production.

and made the change in release.sgml. Do we need more?

I don't much like "might have more bugs". Perhaps "might be less robust"
or "might be less stable"?

cheers

andrew

Andrew Dunstan wrote:

Bruce Momjian wrote:

I have shortened your paragraph to:

Because Win32 is significantly different from the Unix platforms
supported in previous releases, this port might have more bugs
than other supported platforms in this release. Please
test it thoroughly before using it in production.

and made the change in release.sgml. Do we need more?

I don't much like "might have more bugs". Perhaps "might be less robust"
or "might be less stable"?

"lacks the extended testing"? Or even "we don't have an extended
track-record on this platform".

--
Richard Huxton
Archonet Ltd

On Thu, 26 Aug 2004, Richard Huxton wrote:

"lacks the extended testing"?

That's good. We don't need to focus on bugs. The word bug and postgresql
should never occur in the same sentence... hmm...

--
/Dennis Bj�rklund

OK, new wording:

Because Win32 is significantly different from the Unix platforms
supported in previous releases, it has not been tested as
extensively as other supported platforms in this release. Please
test it thoroughly before using it in production.

---------------------------------------------------------------------------

Dennis Bjorklund wrote:

On Thu, 26 Aug 2004, Richard Huxton wrote:

"lacks the extended testing"?

That's good. We don't need to focus on bugs. The word bug and postgresql
should never occur in the same sentence... hmm...

--
/Dennis Bj?rklund

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

After a long battle with technology, andrew@dunslane.net (Andrew Dunstan), an earthling, wrote:

Bruce Momjian wrote:

I have shortened your paragraph to:

Because Win32 is significantly different from the Unix platforms
supported in previous releases, this port might have more bugs
than other supported platforms in this release. Please
test it thoroughly before using it in production.

and made the change in release.sgml. Do we need more?

I don't much like "might have more bugs". Perhaps "might be less
robust" or "might be less stable"?

How about:

"We cannot be as confident in its stability as we are for other
platforms that have been tested and supported across numerous past
releases."
--
(reverse (concatenate 'string "gro.mca" "@" "enworbbc"))
http://cbbrowne.com/info/rdbms.html
Q: What does the function NULL do?
A: The function NULL tests whether or not its argument is NIL or not. If
its argument is NIL the value of NULL is NIL.
-- Ken Tracton, Programmer's Guide to Lisp, page 73.