> On 22 Oct 2025, at 01:56, David Rowley <dgrowleyml@gmail.com> wrote:
>
> On Wed, 22 Oct 2025 at 01:09, Daniel Gustafsson <daniel@yesql.se> wrote:
>> The featurematrix was a rabbithole that went deep, looking closer I realized
>> that links were using a variety of target versions: /current/, /devel/, or the
>> version which was current at the time of adding. This means that for old
>> features we we still linking to the 9.x version of the docs, which clearly
>> isn't helpful to our users. We also had quite a few /static/ links which 301
>> redirects since we don't have that level anymore.
>>
>> On top of that, a few links in the descriptions were not markdown encoded and
>> thus rendered as plain text.
>>
>> The attached v3 fixes these in 0003 and 0004.
>
> * The new link seems dead. Looks like it should be
> https://www.postgresql.org/docs/current/brin.html#BRIN-INTRO
Indeed, fixed.
> * Looks like this will cause issues as "current" moves forward in
> versions and we add new sections before this:
>
> - url: https://www.postgresql.org/docs/16/functions-aggregate.html#id-1.5.8.27.5.2.4.1.1.1.1
> + url: https://www.postgresql.org/docs/current/functions-aggregate.html#id-1.5.8.27.6.2.4.1.1.1.1
>
> If I adjust /current/ to /17/ it works, but not before then. Is there
> anything more future proof we can do there?
Adding anchors with indexterms, as you mention downthread, to anything in the
featurematrix is the solution to providing stable links. A manual step of
verifying links at time of GA is another option, which is probably a good idea
regardless.
> Going back to /14/ that links to left() instead of casefold().
Right, but that's only if you edit the url and I don't think we can ever
protect ourselves from users manually testing links. If you go to the above
page and click on the link for v14 documentation it won't use the anchor and
thus works.
> * The following no longer links to where it intended to:
>
> - url: https://www.postgresql.org/docs/current/static/collation.html#id-1.6.10.4.5.7.5
> + url: https://www.postgresql.org/docs/current/collation.html#id-1.6.10.4.5.7.5
>
> i.e basically someone made the same mistake as the above ones were
> about to make.
Interesting, the history of that edit is lost as it was done when the matrix
was defined in the database and not in a standalone yaml file. The correct
link seems to be:
https://www.postgresql.org/docs/current/collation.html#ICU-CUSTOM-COLLATIONS
--
Daniel Gustafsson
