Обсуждение: Year of first commit

Marcos Pegoraro <marcos@f10.com.br> writes:
> Today Amit Langote committed execScan.h and put on its header years from
> 1996 to 2025. Those years are what exactly ?

1996 is when the open-source Postgres effort started.

There's a bigger philosophical issue here, which is whether a new file
shouldn't just list the current year as the initial copyright year.
In this case, a lot of the code in execScan.h was formerly in
execScan.c, so copying its copyright notice seems like clearly the
right thing.  We've generally taken the attitude that most stuff in
Postgres has pretty clear traceability to older work, so we prefer
to use the same copyright dates as related older files even in less
clear-cut cases.

            regards, tom lane

Marcos Pegoraro <marcos@f10.com.br> writes:
> But these numbers seem inaccurate, because there are 1644 copyright strings
> but 1521 are 1996-2025.

[ shrug... ]  I did not claim that this idea has been adhered to 100%.
In a code base as old and large as ours, very few things are adhered
to 100%.

            regards, tom lane

On 2025-Jan-21, Marcos Pegoraro wrote:

> But these numbers seem inaccurate, because there are 1644 copyright strings
> but 1521 are 1996-2025. So, JSONB, BRIN, FDW, Logical Replication and lots
> of others, all of them were started in 1996 or have any related code ?

I have to ask, why do you think this is important?

For instance, I invented the BRIN interface in 2014 mostly out of thin
air.  Commit 7516f5259411 (its first) added a lot of new files, and
almost all of them were marked with (c) 1996-2014, and all of them (save
one, probably by accident) had a (c) 1996 Regents of UC line in
addition.  The reason I did that, is that most of those files (or
probably all of them) were copying something from elsewhere in the tree.
Was it worth my time, figuring out exactly which files didn't have a
single line copied from elsewhere so that I could restrict the date?  My
analysis at the time was that it wasn't.  Because, you know what?  It is
unlikely to be important.

There was a big licensing change around year 2000 or 2001 from what was
believed to be the original Berkeley license to the license we use today
(widely known as the PostgreSQL license).  The new license is pretty
liberal, to the point that there's no reason to try to establish who
owns the copyright for each file.  Under the old license it might have
been important to determine ownership of individual files.

-- 
Álvaro Herrera        Breisgau, Deutschland  —  https://www.EnterpriseDB.com/
Maybe there's lots of data loss but the records of data loss are also lost.
(Lincoln Yeoh)

Marcos Pegoraro <marcos@f10.com.br> writes:
> Why is this important ? Well, if that file has a 2014-2025, I can know how
> old it is, more or less which version was included, etc
> Additionally, I see all of you adding or removing a single letter to
> sources, why not adjust those years on header files ?

Our git history is a far more reliable guide to how long a specific
file has existed or which versions included it.  The copyright notices
exist only for legal reasons.  In my layman's understanding of
copyright law, it's better to take a maximalist approach to labeling
the copyright years.  That's why, for example, we advance the end
date on every file every year, even if it receives no other changes
that year.

            regards, tom lane