Обсуждение: Final cleanup of SQL:1999 references

This doc patch replaces all inappropriate references to SQL:1999 when it
is used as if it were the latest (and/or still valid) SQL standard.

SQL:2003 is used in its place.

Best Regards, Simon Riggs

Simon Riggs wrote:
> This doc patch replaces all inappropriate references to SQL:1999 when it
> is used as if it were the latest (and/or still valid) SQL standard.

Applied, thanks.

-Neil

Am Mittwoch, 13. Juli 2005 18:01 schrieb Simon Riggs:
> This doc patch replaces all inappropriate references to SQL:1999 when it
> is used as if it were the latest (and/or still valid) SQL standard.
>
> SQL:2003 is used in its place.

I don't necessarily consider this search and replace to be appropriate.  All
information for users looking for SQL:1999 compliance information is now
lost.  Also, phrases like "This is conforming to SQL:2003." give the wrong
impression that it is not conforming to SQL:1999.  I think it would be less
confusing in these cases to simply write "This is conforming to the SQL
standard." and then mention in the appendix that we consider SQL:2003 to be
the baseline.

--
Peter Eisentraut
http://developer.postgresql.org/~petere/

Peter Eisentraut wrote:
> I think it would be less confusing in these cases to simply write
> "This is conforming to the SQL standard." and then mention in the
> appendix that we consider SQL:2003 to be the baseline.

How would this help? ISTM you are just suggesting we replace "conforming
to SQL:2003" with "conforming with the SQL standard", and a note in the
appendix that indicates by "SQL standard" we actually mean "SQL:2003".
If people are really concerned about whether a given feature conforms to
SQL-92, SQL:1999, or SQL:2003, all we have done is provided them with
the same information in a slightly different form.

-Neil

Am Donnerstag, 14. Juli 2005 15:53 schrieb Neil Conway:
> How would this help? ISTM you are just suggesting we replace "conforming
> to SQL:2003" with "conforming with the SQL standard", and a note in the
> appendix that indicates by "SQL standard" we actually mean "SQL:2003".

I recall that most mentions of "SQL:1999" were specially not spelled "SQL
standard" because the feature was not present in earlier versions.  That is,
we gave the version number only when it was important.  Now by replacing that
with "SQL:2003" leads ad absurdum and should just be replaced with "SQL
standard".

Unless you also plan to replace all mentions of "SQL standard" with
"SQL:2003".

--
Peter Eisentraut
http://developer.postgresql.org/~petere/

Neil Conway <neilc@samurai.com> writes:
> If people are really concerned about whether a given feature conforms to
> SQL-92, SQL:1999, or SQL:2003, all we have done is provided them with
> the same information in a slightly different form.

No, you have *removed* the information.  The convention we were
effectively following was that a reference to "SQL-xxxx" rather than
just "SQL" implies that xxxx was the first version to say that.

I agree with Peter that a search-and-replace patch is entirely
off the mark.  Please revert it and do some research instead.

            regards, tom lane

On Thu, 2005-07-14 at 10:12 -0400, Tom Lane wrote:
> Neil Conway <neilc@samurai.com> writes:
> > If people are really concerned about whether a given feature conforms to
> > SQL-92, SQL:1999, or SQL:2003, all we have done is provided them with
> > the same information in a slightly different form.
>
> No, you have *removed* the information.  The convention we were
> effectively following was that a reference to "SQL-xxxx" rather than
> just "SQL" implies that xxxx was the first version to say that.
>
> I agree with Peter that a search-and-replace patch is entirely
> off the mark.  Please revert it and do some research instead.

I note that nobody asked whether any research was conducted in support
of the patch, you have just assumed there was none. I'm not clear why
anybody would reach that conclusion; if there is another issue here, I'm
happy to discuss that on or off-list.

So that we are clear, I referred to the current standard guide for each
change. Any discussion on this patch should be on a point by point
basis, not just on "the whole thing is wrong" basis because we have used
the emotional phrase "search-and-replace" to refer to all of the
proposed changes.

The main point is that SQL:1999 no longer has any validity as a standard
and has been wholly superceded by SQL:2003. SQL:1999 has interest only
for historical reasons, for those who care when a particular feature was
introduced.

In some paragraphs there was discussion of historical aspects. I believe
I have included an appropriate statement for each. In most cases there
is no historical discussion on whether particular features were
standardised at particular times, so in those cases I have not sought to
include that information.

On Thu, 2005-07-14 at 10:12 -0400, Tom Lane wrote:
> The convention we were
> effectively following was that a reference to "SQL-xxxx" rather than
> just "SQL" implies that xxxx was the first version to say that.

If that was the convention, it is not clear in the docs. They simply look
like they have not been properly updated to reflect the new version of the
standard. That was my objective.

Peter's suggestion of just saying "the SQL standard" seems motivated by
reducing the amount of rework for translators. I would support that, but
ask that Peter make those changes and reapply.

Best Regards, Simon Riggs

Simon Riggs wrote:
> The main point is that SQL:1999 no longer has any validity as a standard
> and has been wholly superceded by SQL:2003. SQL:1999 has interest only
> for historical reasons, for those who care when a particular feature was
> introduced.

Right; I guess the question is whether we should attempt to cater to the
latter group. Personally I think most users are only concerned with
whether a given feature conforms to the most recent version of the
standard. Including a haphazard mix of SQL-92, SQL:1999, and SQL:2003
just leads to confusion (if Simon didn't notice this convention, it is a
fair bet not many users did, either). If people are actually concerned
about what version of the standard introduced a particular feature, they
are better, more authoritative sources with this information (e.g. the
standards themselves).

There is also the separate issue of whether we should refer to SQL:2003
or "the SQL standard". On second thought, I'm happy with the latter.

-Neil

Some time ago, Peter Eisentraut <peter_e@gmx.net> wrote:
> Am Mittwoch, 13. Juli 2005 18:01 schrieb Simon Riggs:
>> This doc patch replaces all inappropriate references to SQL:1999 when it
>> is used as if it were the latest (and/or still valid) SQL standard.
>>
>> SQL:2003 is used in its place.

> I don't necessarily consider this search and replace to be appropriate.  All
> information for users looking for SQL:1999 compliance information is now
> lost.

Also, there are places where it completely destroyed the meaning, as in
http://developer.postgresql.org/docs/postgres/sql-select.html#AEN48034
(the "Namespace Available to GROUP BY and ORDER BY" subsection), where
the intended distinction between SQL-92 and later versions of the spec
has been obliterated, and the text is now actually self-contradictory.

> Also, phrases like "This is conforming to SQL:2003." give the wrong
> impression that it is not conforming to SQL:1999.  I think it would be less
> confusing in these cases to simply write "This is conforming to the SQL
> standard." and then mention in the appendix that we consider SQL:2003 to be
> the baseline.

I agree: we should just say SQL except where there is an intention to
distinguish different versions of the spec, and in that case mention
the earliest spec version for which the particular statement is true.
This convention will not require any future search-and-replaces.

Barring some fairly convincing objections, I am going to undo most of
this patch later this week, and instead do it as Peter suggests.

            regards, tom lane

On Tue, 2005-11-01 at 14:43 -0500, Tom Lane wrote:
> Some time ago, Peter Eisentraut <peter_e@gmx.net> wrote:
> > Am Mittwoch, 13. Juli 2005 18:01 schrieb Simon Riggs:
> >> This doc patch replaces all inappropriate references to SQL:1999 when it
> >> is used as if it were the latest (and/or still valid) SQL standard.
> >>
> >> SQL:2003 is used in its place.

...

> > Also, phrases like "This is conforming to SQL:2003." give the wrong
> > impression that it is not conforming to SQL:1999.  I think it would be less
> > confusing in these cases to simply write "This is conforming to the SQL
> > standard." and then mention in the appendix that we consider SQL:2003 to be
> > the baseline.
>
> I agree: we should just say SQL except where there is an intention to
> distinguish different versions of the spec, and in that case mention
> the earliest spec version for which the particular statement is true.
> This convention will not require any future search-and-replaces.
>
> Barring some fairly convincing objections, I am going to undo most of
> this patch later this week, and instead do it as Peter suggests.

Agreed. No need for future changes is a convincing argument.

Should we make the phrase "SQL Standard" an xref to the appropriate
section in the docs? That would help with any further confusion on this.

Best Regards, Simon Riggs