Search engine linking to latest docs

On 7/25/22 6:08 PM, Trevor Gross wrote:

Hey all,

I think the PG community is pretty aware that there are some issues with
search engines taking users to the right versions of the docs. You
google something simple like "postgres drop column" and the first
postgresql.org <http://postgresql.org&gt; result takes you to the v7.4 docs
- EOL over a decade ago. It's a similar story for many searches, lots of
results take you to v8 or v9 docs first.

One of the efforts this year to improve it has already yielded results,
where much more significant search traffic going to the latest docs
(stats here):

/messages/by-id/25aa516b-4fa7-5083-366e-09cf4f838e0f@postgresql.org

The 7.4 page does include the canonical link, but perhaps due to the
amount of links going to the 7.4 ALTER TABLE page / the term not having
been updated in awhile, it has not been reindex yet.

(Perhaps these recent searches will help with that reindexing :)

(Note that if you've clicked "current" in the docs before, it will
automatically redirect you. This is more relevant to newer users, and
the actual links on google).

I do not have the personal expertise to help this situation at all, but
I would like to call attention to this active thread for the rust docs
repo - https://github.com/rust-lang/docs.rs/issues/1438
<https://github.com/rust-lang/docs.rs/issues/1438&gt;. The corresponding
site docs.rs <http://docs.rs&gt; has similar issues of always linking to
outdated versions, and they're currently in the process of working to
improve things. The thread contains a good bit of findings and
trial/error. Knowing that PG has similar issues, it seems like there is
a possibility of benefiting from mutual lessons learned in improving SEO
and linking.

Agree in mutual benefits. This topic has been worked on for decades(!)
now on the pgsql-www mailing list. For instance, our findings on
"rel=canonical" are the opposite of what appears in that GitHub thread,
and matches what some other OSS projects are doing (e.g. Django).

We consider "latest" to be a "superset" of the other versions. We
explicitly do not want to redirect from old to new, given we support 5
(and briefly 6) major PostgreSQL versions at any given time and want to
ensure users have access to older doc versions.

Again, I have no concrete proposals to make, but wanted to draw
attention to the issue and a good thread of ways to improve / possible
solutions.

Thanks,

Jonathan

This topic has been worked on for decades(!)

now on the pgsql-www mailing list

I thought I opened a new thread on what I thought was a newish topic, but I
guess it turns out I was just searching the wrong mailing list. Thanks for
the insights, I’ll consider this closed :)

On Mon, Jul 25, 2022 at 18:18 Jonathan S. Katz <jkatz@postgresql.org> wrote: