Outdated Download ?
(1) By Arend-Jan Westhoff (Arend-Jan) on 2024-01-30 14:05:54 [link] [source]
(2) By Richard Hipp (drh) on 2024-01-30 14:27:09 in reply to 1 [link] [source]
Your links are pointing to one of the backup Fossil sites, not the main site. You are pointing to
Notice the "2" after "www". The main site is:
The problem was that the Download page, which is stored as unversioned files, was not getting synced up correctly with the main website. I had to run this command from my desktop:
fossil sync -u -all
And that seems to have cleared the problem. I suppose I need to update the release procedures to remind me to do that in the future.
Thanks for bringing the problem to my attention.
(3.1) By Warren Young (wyoung) on 2024-01-30 21:34:24 edited from 3.0 in reply to 2 [source]
For some reason, the www2 and www3 servers are getting more Google juice, at least sometimes, causing them to come up ahead of www links. Canonicalization would fix this, but there isn't a static string you can insert into the page header for this.
Fossil can generate the per-page dynamic URL needed for this, then inject it via a <link rel=canonical> tag into every header emitted by src/style.c. It does have the info needed to produce that: the base URL from the existing "Canonical Server URL" setting (a.k.a. email-url), and the rest from info given in the HTTP GET request, used to dispatch the call.
The main worry I have about that is whether email-url copies down during clone, but it looks like it didn't do so, here at least. My testing mirror should not go around announcing that its canonical URL is https://fossil-scm.org/home/path/to/resource on every page in Fossil UI; that implies that the content is the same, but that isn't necessarily so, as currently when the default skin is all different, on purpose.
Secondarily, it's a small overreach to reuse a setting meant for email alerts, but since I can't justify creating a new one or renaming this one to be more general, it'll have to do.
An alternate plan is to do it at the HTTP proxy level, althttpd in this case. It can put a "Link" header into every response to tell search engines they shouldn't be keeping the duplicate www2 and www3 results.
Stale mirrors aside, there's no point diluting our Google juice by giving search engines the false impression that there are three competing sites with the same content.