Re: [PATCH] Allow Postgres to pick an unused port to listen

Alvaro, Tom,

On Mon, May 8, 2023 at 4:49 PM Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> Alvaro Herrera <alvherre(at)alvh(dot)no-ip(dot)org> writes:
> > This made me wonder if storing the unadorned port number is really the
> > best way. Suppose we did extend things so that we listen on different
> > ports on different interfaces; how would this scheme work at all?
>
> Yeah, the probability that that will happen someday is one of the
> things bothering me about this proposal. I'd rather change the
> file format to support that first (it can be dummy for now, with
> all lines showing the same port), and then document it second.
>

How soon do you think the change will occur that will allow for choosing
different ports on different interfaces? I am happy to help address this.

Relying on a variable number of lines may be counter-productive here if we
want postmaster.pid to be easily readable by shell scripts. What if we
improved the port line to be something like this?

```
127.0.0.1=5432 ::1=54321
```

Basically, a space-delimited set of address/port pairs (delimited by `=` to
allow IPv6 addresses to use a colon). If we allow the address side to be
dropped, the current format (`5432`) will also be correct parsing-wise.

--
Y.