Complete the galaxyproject.eu → hub redirect map#2198
Open
itisAliRH wants to merge 7 commits into
Open
Conversation
Extend the galaxyproject.eu server block with the full legacy URL to galaxyproject.org redirect table derived from the EU website migration audit manifest, completing the work started in #421. All 1308 routable legacy records are covered: 574 news posts and 258 events get exact `location =` rules pointing at their verified hub page, subdomain landing pages, core EU pages and member subsites are handled by deterministic regex rules, and un-migrated content (chiefly the ~222 autogenerated tool-update posts) collapses to the news index per the redirect table in galaxyproject/galaxy-hub#1480. Exact blocks rather than a single regex transform are required because the hub does not normalise slugs -- uppercase, camelCase, underscore and re-dated event URLs all 404 -- and stock nginx cannot lower-case in a rewrite. Regexes that interpolate a capture into their target are therefore case-sensitive; the rest are case-insensitive and tolerant of a trailing slash and of the legacy .html suffix, following #2108. Two rewrites fold the plain.html, -plain and event.ics variants onto the canonical path so one rule per item suffices. Unmatched paths now land on /eu/ instead of being proxied to usegalaxy-eu.github.io. Verified with nginx -t and by replaying all 5613 legacy URLs against a local nginx: every one returns 301, and all 906 distinct targets resolve on galaxyproject.org (zero 404s).
The audit manifest marks these 89 posts and 44 events as `migrated` and the hub content exists and serves 200, but no exact `location =` rule was generated for them, so they fell through to the blanket index redirect (/news/ or /eu/events/) instead of reaching their article. Targets are lower-cased: the hub lower-cases published slugs. Three manifest entries were mis-cased and 404 as recorded -- /events/2022-11-BH2022/ and /events/2022-10-PSU-Workshop/ -- their lower-cased forms are 200. All 133 targets were verified live before being written. Exact rules: 832 -> 965 (posts 574 -> 663, events 258 -> 302).
Every exact `location =` block is keyed on a trailing-slash URL, but the legacy host also answers the slashless form: GitHub Pages 301s /posts/2018/10/08/gxadmin -> /posts/2018/10/08/gxadmin/, so slashless URLs are real, linked and working in the wild. A slashless request matched no exact block and fell through to the blanket index redirect, silently bypassing the entire 965-rule table. Two server-level rewrites fold the slashless and .html forms onto the canonical key, following the existing plain/ics folding pattern. `[^/]+?` cannot match a trailing slash, so a rewritten URI cannot re-enter these rules and no rewrite loop is possible. The .html form 404s on the legacy host today (posts and events are directory -style URLs); folding it is defensive and can only upgrade a dead URL. Verified against a live nginx: slashless, trailing-slash, .html, -plain, /plain.html and /event.ics variants all resolve to the article, dotted slugs (Galaxy-17.09) survive normalisation, and unmigrated tool-update posts still reach the blanket rule.
The overrides for index-microbiome, index-ssh and index-test were keyed only on the .html and trailing-slash forms, so the BARE form -- which is the canonical published URL (/index-ssh 200, /index-ssh/ 404 on the legacy host) -- fell through to the generic index- rule and was sent to /bare/eu/usegalaxy/<name>/, which 404s on the hub. All three were broken-target. Replaces the six exact blocks (two of which were dead: the trailing-slash URLs were never published) with three regex blocks covering the bare, .html and trailing-slash forms in one place, ordered ahead of the generic rule. index-microbiome.md is a symlink to index-metagenomics.md in the legacy repo, so it is the metagenomics page; content/redirects.yaml in the hub independently declares the same target. Verified 200. index-ssh and index-test have no hub counterpart -- index-ssh does not even exist in the legacy source tree, only in the deployed branch -- so they point at the EU landing page and are recorded as content gaps.
All 467 legacy asset URLs hit the catch-all and 301 to /eu/, so an image request is answered with a 52 KB HTML page at HTTP 200. That is strictly worse than a 404: no client or crawler can detect the breakage, and every legacy hotlink -- in third-party pages, archived posts and Galaxy tool help text -- renders as a broken image. 163 of them exist on the hub at the identical path, so a single prefix rule recovers them (/assets/media/tk_besoek.jpg -> 200 image/jpeg, verified). The remaining ~290 now 404 honestly. css/, fonts/ and js/ are deliberately included rather than carved out: none exist on the hub, but a 301 that ends in a 404 is a truthful answer whereas the catch-all returns 200 text/html for a stylesheet.
Feeds first -- these are subscription endpoints, and a 301 to an HTML page
breaks every reader silently:
/feed.xml -> /eu/feed.atom (jekyll-feed site-wide posts feed;
RSS -> Atom, readers follow the 301)
/events-ical -> /eu/events/calendar.ics (legacy served this as text/html, so
strict clients never worked; the hub
serves real text/calendar)
Per-event /event/<slug>/event.ics keeps folding to the event page rather than
to the EU calendar: that calendar holds only the 12 upcoming events, so a
request for a 2022 event would import unrelated ones.
Pages that already have hub content but were being swallowed by a fallback:
/galaxy/data-policy -> /eu/storage/ (content/eu/common/data-policy.md is an
include fragment; the hub publishes it
under /eu/storage/)
/people -> /eu/people/
{7 subsites}/people -> /<sub>/people/
{7 subsites}/publications -> /<sub>/citations/ (renamed on the hub)
/freiburg/{projects,services}
/freiburg/jobs[/*] -> /careers/
/freiburg/jobs is the one URL where this map was worse than the status quo:
it 301s to a working page today, but the subsite fallback would have sent the
three job ads to the Freiburg home page.
/livestream, /livestream-plain and /wall/* have no hub equivalent and are
recorded explicitly so a later reader can see they were considered.
All 30 targets verified 200 live.
The SSH (Social Sciences and Humanities) front page has been ported to the hub at /bare/eu/usegalaxy/ssh/, so /index-ssh no longer needs the /eu/ fallback. All three forms (bare, .html, trailing slash) follow. /index-test keeps the /eu/ fallback and is now commented as an intentional retirement: it was a throwaway test-server page and has no hub counterpart by decision, so the fallback is permanent rather than a gap awaiting a migration. Refs usegalaxy-eu/issues#943.
bgruening
reviewed
Jul 23, 2026
| # /feed.xml was jekyll-feed's site-wide posts feed; /eu/feed.atom is the hub | ||
| # equivalent. RSS -> Atom is a format change, but every mainstream reader | ||
| # follows the 301 and parses Atom. | ||
| location ~* "^/feed\.xml$" { |
Member
There was a problem hiding this comment.
does that really exist? https://usegalaxy-eu.github.io/feed.xml ?
|
|
||
| # The rest should continue to be served by the European website, but under the github.io domain. | ||
| return 301 https://usegalaxy-eu.github.io$request_uri; | ||
| # No hub equivalent exists for these; recorded explicitly so that a future |
| location ~* "^/livestream(?:-plain)?(?:\.html|/)?$" { | ||
| return 301 https://galaxyproject.org/eu/; | ||
| } | ||
| location ~* "^/wall(?:/.*)?$" { |
Member
There was a problem hiding this comment.
drop for now and ask @wm75 we should have a mastodon wall on the Hub already.
Member
There was a problem hiding this comment.
Yes, there is https://galaxyproject.org/wall/ already - we just have to decide how much of this we want to host ourselves. Currently, https://gxy.io/fediwall points to fediwall.social passing it the config file at https://raw.githubusercontent.com/bgruening/fediwall-config/refs/heads/main/fediwall-config.json as a param.
I guess we could decide to:
- host just that config file
- host a fediwall instance (https://github.com/defnull/fediwall; static site) and its config file ourselves
If this takes some time to decide, we could just put a redirect to https://gxy.io/fediwall here for now.
Member
There was a problem hiding this comment.
Thanks @wm75
@itisAliRH please redirect this one then to https://galaxyproject.org/wall/ for the time being.
| return 301 https://galaxyproject.org/eu/; | ||
| } | ||
|
|
||
| # --------------------------------------------------------------------------- |
Member
There was a problem hiding this comment.
I think we should remove that and keep the config small.
We made sure that the posts / events are redirected on the Hub website. And if we add as a last resort fallback to the old jekyll website a info page (this website has migrated, please visit ...) that should be fine.
| location = /event/2022-10-28-analysis-data-Galaxy/ { return 301 https://galaxyproject.org/events/2022-10-psu-workshop/; } | ||
| location = /event/2022-10-28-community-call/ { return 301 https://galaxyproject.org/events/2022-11-17-community-call/; } | ||
| location = /event/2022-10-28-single-cell-RNA-seq/ { return 301 https://galaxyproject.org/events/2023-02-ebi-single-cell/; } | ||
| location = /event/2022-10-31-Galaxy-biohackaton-Europe/ { return 301 https://galaxyproject.org/events/2022-11-bh2022/; } |
Member
There was a problem hiding this comment.
drop everything until here, all posts and events
| location = /event/2022-10-28-single-cell-RNA-seq/ { return 301 https://galaxyproject.org/events/2023-02-ebi-single-cell/; } | ||
| location = /event/2022-10-31-Galaxy-biohackaton-Europe/ { return 301 https://galaxyproject.org/events/2022-11-bh2022/; } | ||
|
|
||
| # Static assets keep their path on the hub, so map the prefix straight across. |
| return 301 https://galaxyproject.org$request_uri; | ||
| } | ||
|
|
||
| # Blanket rules for content that was never migrated individually -- chiefly the |
| location ~* "^/posts/\d{4}/\d{2}/\d{2}/" { return 301 https://galaxyproject.org/news/; } | ||
| location ~* "^/event/" { return 301 https://galaxyproject.org/eu/events/; } | ||
|
|
||
| # Anything not matched above is legacy content with no hub equivalent. |
Member
There was a problem hiding this comment.
I would drop this as well, or redirect to a speical site that is:
- nice looking
- Apologize that the link is not working
- redirecting users to news, events, storage etc ....
Member
|
@itisAliRH this now needs a rebase. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Replaces the blanket
galaxyproject.eu→usegalaxy-eu.github.ioshim with a per-URL 301 map ontogalaxyproject.org, derived from the migration audit manifest and verified URL-by-URL against the live hub.master's config currently ends with:So today the entire legacy long tail is served by the GitHub Pages build of
usegalaxy-eu/website— the very repo that is slated for archival. This branch removes bothgithub.ioreferences (0 remain).Deploying this map is therefore a precondition for archiving
usegalaxy-eu/website, not a follow-up to it. Archiving that repo first would break every legacy URL, including the ~222 tool-update posts and every asset. This ordering constraint was not previously recorded anywhere in the migration work.Rule counts
location =location ~location ^~rewriteExact blocks: 663 news posts + 302 events + 4 structural.
Coverage
Posts and events (1187 published legacy URLs)
full(specific rule + hub content + live 200)redirect-only(blanket index)broken-targetno-ruleThe remaining 222 are exactly the auto-generated
tool-updateposts, collapsed to/news/by design (galaxyproject/galaxy-hub/issues/1480). Every other post and event now resolves to its own article.Pages, subdomains, subsites, feeds, assets (615 legacy URLs)
fullredirect-onlyno-rulebroken-targethonest-404(asset absent from hub, deliberate)Slashless URL normalisation — the largest defect fixed here
Every exact block is keyed on a trailing-slash URL, but the legacy host also answers the slashless form (GitHub Pages 301s
/posts/2018/10/08/gxadmin→/posts/2018/10/08/gxadmin/), so slashless URLs are real, linked and working in the wild. Without normalisation they matched no exact block and fell through to the blanket index — silently bypassing theentire rule table.
Measured by replaying all 1187 post/event URLs in slashless and
.htmlform against a real nginx running each config:.htmlvariant requests (2374)The 444 remaining are exactly 2 × the 222 tool-update posts.
What changed
migratedand the hub content exists and serves 200, but no rule was ever generated, so they landed on the index. Three manifest targets were mis-cased and 404 as recorded (/events/2022-11-BH2022/,/events/2022-10-PSU-Workshop/); the hub lower-cases published slugs, so all targets are written lower-cased and were re-verified live..htmlnormalisation — two server-level rewrites folding both forms onto the canonical key, following the existing plain/ICS folding pattern.[^/]+?cannot match a trailing slash, so no rewrite loop is possible.index-*bare forms./index-microbiome,/index-sshand/index-testwere keyed only on.htmland trailing slash, so the canonical bare form hit the generic rule and 404ed — all three werebroken-target. ** Six exact blocks (two of them dead) are replaced by three regex blocks covering all forms.index-microbiome.mdis a symlink toindex-metagenomics.mdin the legacy repo, andcontent/redirects.yamlin the hub independently declares the same target, so it now points at/bare/eu/usegalaxy/metagenomics/./assets/pass-through. All 467 legacy asset URLs previously hit the catch-all and 301ed to/eu/, answering an image request with a 52 KB HTML page at HTTP 200 — strictly worse than a 404, because nothing can detect the breakage. 163 exist on the hub at the identical path and are now recovered; the rest 404 honestly.The asset decision
css/,fonts/andjs/are deliberately included rather than carved out. None of them exist on the hub (it ships its own theme chrome), so they will 404 — but a 301 that ends in a 404 is a truthful answer, whereas the catch-all returns200 text/htmlfor a stylesheet request. A carve-out would add rules to produce a strictly worse outcome.Feeds
These are live subscriber endpoints, so a 301 to an HTML page breaks every reader silently.
/feed.xml/eu/feed.atom/events-ical/eu/events/calendar.icstext/html, so strict clients never worked; the hub serves realtext/calendar. Matchescontent/redirects.yaml./event/<slug>/event.icskeeps folding to its event page rather than to the EU calendar: that calendar contains only the 12 upcoming events, so a request for a 2022 event would import unrelated ones.Known gaps
These are deliberate and verified, not oversights.
/news/. Auto-generated changelog posts, collapsed by design (Hub unification: Initial launch checklist galaxyproject/galaxy-hub#1480).media/images + 10css/, 9fonts/, 4js/). These now 404 honestly instead of returning an HTML page at 200. Recovering them requires migrating the files, not a redirect change./index-sshand/index-testhave no hub counterpart →/eu/.index-sshis stranger than it looks: it returns 200 on the deployed legacy site but does not exist in the source tree at all — the deployed GitHub Pages branch carries files absent frommaster. Worth a separate look, since other deployed-only files may exist that neither a source derivation nor a crawl can see./<sub>/galaxy/,/<sub>/galaxy/gxnews,/<sub>/galaxy/gxeventsfor 6 subsites, plus/freiburg/maintenance/) → the subsite home./bare/<sub>/news/and/bare/<sub>/events/were probed and 404 on the hub; there is nothing to point at./livestream,/livestream-plain,/wall/,/wall/social_template→/eu/. No hub equivalent. Given explicit rules anyway so a future reader can see they were considered rather than missed./error_pages/*stay on the catch-all — infrastructure, not user-facing./_bibliography/*.bibneed no rule: an underscore directory Jekyll never published (404 on legacy today). The source data was migrated tocontent/*/citations/*.bib.Validation
nginx -tpasses on the rendered config..htmlvariants replayed through a real nginx: 0 disagreements with the canonical form.Note for reviewing on macOS
nginx -tfails locally on macOS withduplicate location "/event/2022-07-05-community-call/". This is a platform artifact, not a config defect: nginx definesNGX_HAVE_CASELESS_FILESYSTEMon Darwin and Win32, songx_filename_cmpcompares location keys case-insensitively there. Linux — the deploy target — is case-sensitive._events/2022-07-05-Community-Call.mdand_events/2022-07-05-community-call.mdare two distinct source files with distinct content, both live 200 on the legacy site, mapping to two different hub events. Both rules are required. Verified: 0 duplicates under case-sensitive (Linux) semantics.