From: | Stephen Frost <sfrost(at)snowman(dot)net> |
---|---|
To: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
Cc: | Heikki Linnakangas <hlinnaka(at)iki(dot)fi>, Ian Lawrence Barwick <barwick(at)gmail(dot)com>, pgsql-hackers <pgsql-hackers(at)postgresql(dot)org> |
Subject: | Re: [doc] remove reference to pg_dump pre-8.1 switch behaviour |
Date: | 2020-10-23 14:12:11 |
Message-ID: | 20201023141211.GH16415@tamriel.snowman.net |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Greetings,
* Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
> Stephen Frost <sfrost(at)snowman(dot)net> writes:
> > Isn't this a bit pre-mature as we still support running pg_dump against
> > 8.0 clusters..?
>
> The removed para was discussing the behavior of pg_dump itself. What
> server version you run it against isn't relevant.
Ah, alright, that makes a bit more sense then..
> Having said that, there are a *lot* of past-their-sell-by-date bits
> of info throughout our documentation, because we don't have any sort
> of policy or mechanism for getting rid of this kind of backwards
> compatibility note. Maybe we should first try to agree on a policy
> for when it's okay to remove such info.
I would have thought the general policy would be "match what the tool
works with", so if we've got references to things about how pg_dump
works against older-than-8.0 then we should clearly remove those as
pg_dump no londer will run against versions that old.
Extending that to more general notes would probably make sense though.
That is- we'll keep anything relevant to the oldest version that pg_dump
runs against (since I'm pretty sure pg_dump's compatibility goes the
farthest back of anything we've got in core and probably always will).
We do need to decide at what point we're going to move forward pg_dump's
oldest server version support. I had thought we would do that with each
top-level major version change (eg: support 8.0+ until we reach 11.0 or
someting), but that doesn't work since we've moved to a single integer
for major versions. Looking at the timeline though:
2016-10-12: 64f3524e2c8deebc02808aa5ebdfa17859473add Removed pre-8.0
2005-01-19: 8.0 released
So, that's about 10 years.
2010-09-20: 9.0 released
Or about 10 years from today, which seems to me to imply we should
probably be considering moving pg_dump forward already. I'm not really
inclined to do this every year as I don't really think it's helpful, but
once every 5 years or so probably makes sense. To be a bit more
specific about my thoughts:
- Move pg_dump up to 9.0 as the required minimum, starting with v14.
- In about 5 years or so, move pg_dump up to minimum of v10.
(clean up all documentation with older references and such too)
If we wanted to be particularly cute about it, we could wait until v15
to drop support for older-than-9.0, and then v20 would remove support
for older-than-10, and then v25 would remove support for
older-than-v15, etc.
Thanks,
Stephen
From | Date | Subject | |
---|---|---|---|
Next Message | Tom Lane | 2020-10-23 14:28:13 | Re: new heapcheck contrib module |
Previous Message | Mark Dilger | 2020-10-23 14:04:04 | Re: new heapcheck contrib module |