Thread

1. Download statistics

   Peter J. Holzer <hjp-pgsql@hjp.at> — 2025-10-25T07:47:00Z

   The discussion about people avoiding .0 releases over in the "Index
   corruption ..." thread made me wonder how the distribution really looks
   like. How many people do install X.0, X.1, etc. for each major version
   X?
   
   Download statistics are of course quite noisy but I think they should at
   least show the trends. Do you have any and would you mind publishing
   them?
   
   My guess is that the downloads will start low for .0 then increase for
   some time (because most people who installed X.y will then upgrade to
   X.y+1, at least if they are using apt or dnf). At some time after the
   next major release it will start to decline again (because people will
   now use the new major release for fresh installs and also start to
   upgrade existing installations) but with a very long tail.
   
           hjp
   
   -- 
      _  | Peter J. Holzer    | Story must make more sense than reality.
   |_|_) |                    |
   | |   | hjp@hjp.at         |    -- Charles Stross, "Creative writing
   __/   | http://www.hjp.at/ |       challenge!"
   

2. Re: Download statistics

   Tom Lane <tgl@sss.pgh.pa.us> — 2025-10-25T14:38:34Z

   "Peter J. Holzer" <hjp-pgsql@hjp.at> writes:
   > The discussion about people avoiding .0 releases over in the "Index
   > corruption ..." thread made me wonder how the distribution really looks
   > like. How many people do install X.0, X.1, etc. for each major version
   > X?
   
   > Download statistics are of course quite noisy but I think they should at
   > least show the trends. Do you have any and would you mind publishing
   > them?
   
   I imagine we have stats for downloads from www.postgresql.org,
   but it's been many years since we thought those were complete
   or even representative.  Nowadays we assume that most people
   consume Postgres via binary packages built by third parties
   (e.g. Linux distros).  Of course, we have zero visibility into
   the download stats for those.
   
   			regards, tom lane
   
   
   
   

3. Re: Download statistics

   Ron Johnson <ronljohnsonjr@gmail.com> — 2025-10-25T14:43:41Z

   On Sat, Oct 25, 2025 at 10:38 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
   
   > "Peter J. Holzer" <hjp-pgsql@hjp.at> writes:
   > > The discussion about people avoiding .0 releases over in the "Index
   > > corruption ..." thread made me wonder how the distribution really looks
   > > like. How many people do install X.0, X.1, etc. for each major version
   > > X?
   >
   > > Download statistics are of course quite noisy but I think they should at
   > > least show the trends. Do you have any and would you mind publishing
   > > them?
   >
   > I imagine we have stats for downloads from www.postgresql.org,
   > but it's been many years since we thought those were complete
   > or even representative.
   >
   
   I download one set of RPMs (PG 14.19, for example) and then scp them to 15
   different servers.  One set of PG 17.6 binaries gets copied to 30 different
   servers.
   
   -- 
   Death to <Redacted>, and butter sauce.
   Don't boil me, I'm still alive.
   <Redacted> lobster!
   

4. Re: Download statistics

   Peter J. Holzer <hjp-pgsql@hjp.at> — 2025-10-25T15:31:30Z

   On 2025-10-25 10:43:41 -0400, Ron Johnson wrote:
   > On Sat, Oct 25, 2025 at 10:38 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
   > 
   >     "Peter J. Holzer" <hjp-pgsql@hjp.at> writes:
   >     > The discussion about people avoiding .0 releases over in the "Index
   >     > corruption ..." thread made me wonder how the distribution really looks
   >     > like. How many people do install X.0, X.1, etc. for each major version
   >     > X?
   > 
   >     > Download statistics are of course quite noisy but I think they should at
   >     > least show the trends. Do you have any and would you mind publishing
   >     > them?
   > 
   >     I imagine we have stats for downloads from www.postgresql.org,
   >     but it's been many years since we thought those were complete
   >     or even representative.
   > 
   > 
   > I download one set of RPMs (PG 14.19, for example) and then scp them to 15
   > different servers.  One set of PG 17.6 binaries gets copied to 30 different
   > servers.
   
   Yes, these are both examples of what I meant by "noisy". There
   are other sources than www.postgresql.org (and apt.postgresql.org,
   download.postgresql.org, etc.) and one download may be used for many
   installations or sometimes none. So the absolute numbers are almost
   meaningless. However, the trends might still be illuminating. Of course
   the patterns might be totally different for those users who just use
   whatever is included with their Linux distribution (for example the 2
   year release cycle of Debian and Ubuntu would cause some shifts) but
   I'm not convinced it would be that different.
   
           hjp
   
   -- 
      _  | Peter J. Holzer    | Story must make more sense than reality.
   |_|_) |                    |
   | |   | hjp@hjp.at         |    -- Charles Stross, "Creative writing
   __/   | http://www.hjp.at/ |       challenge!"