On Thu, Dec 05, 2019 at 07:12:22PM -0600, Ron wrote:
> On 12/5/19 1:01 PM, Tom Lane wrote:
>> It's been considered, and rejected, many times. Aside from the overhead
>> involved, there are too many different ideas of what such dates ought to
>> mean (e.g., what should happen during dump/restore? does a failed
>> transaction update last-modified? etc etc). You can search the
>> project's mailing list archives if you want to read the prior discussions.
>
> All the other RDBMSs seem to have figured it out.
It does not necessarily mean that Postgres has to do it. FWIW, you
can track that using an even trigger for CREATE TABLE or other objects
which inserts the following in a table of your choice for a given
database:
- The timestamp of the transaction.
- The object name.
- Its class ID, say pg_class::regclass for a table, etc.
--
Michael
