fix: injected standard prefixes reference wikidata (#804)

As reported in #798, the "standard prefixes" injected by WDQS do not align with the WIKIBASE_CONCEPT_URI setting introduced in #771, nor do they match the corresponding values for Wikidata. Currently, the WDQS-injected "standard prefixes" (such as wd:) reference http://wikidata/, which is the internal hostname of the wikibase/mediawiki container within the Docker network.

Particularly in the context of federation, it is essential to maintain the "standard prefixes" for referencing Wikidata [1] [2]. To accommodate the local Wikibase instance, prefixes can be set inline, as is currently done on wikibase.cloud [2]. This pull request addresses this and updates the "standard prefixes" to point to Wikidata. Additional context can be found in [3].

[1] https://www.mediawiki.org/wiki/Wikibase/Wikibase.cloud/First_steps#View_your_data_using_the_Query_Service
[2] https://phabricator.wikimedia.org/T335448
[3] https://phabricator.wikimedia.org/T379232

build/wdqs/entrypoint.sh

@@ -13,7 +13,24 @@ done
 
 set -eu
 
-export BLAZEGRAPH_OPTS="${BLAZEGRAPH_EXTRA_OPTS} -DwikibaseHost=${WIKIBASE_HOST}"
+# Options provided to WDQS (blazegraph) when running as query service instance
+#
+# Note: We MUST not provide -DwikibaseHost=${WIKIBASE_HOST} here, otherwise
+# WDQS would re-use the wd: et al. prefixes for the local wikibase instance.
+# This is unintended, especially in the context of federation. wd: prefixes
+# should remain in place for referencing wikidata. The local instance should
+# choose its own prefixes, as described here: 
+# https://www.mediawiki.org/wiki/Wikibase/Wikibase.cloud/First_steps#View_your_data_using_the_Query_Service
+# Some further thoughts on prefixes: https://phabricator.wikimedia.org/T335448
+#
+# In other words, WDQS does not know about the hostname of the wiki it gets
+# its data from. Is is solely the task of the updater to feed data from the
+# wiki into the WDQS instance.
+export BLAZEGRAPH_OPTS="${BLAZEGRAPH_EXTRA_OPTS}"
+
+# Options provided when running as wdqs-updater
+#
+# Here we provide -DwikibaseHost to reference the wiki to poll updates from.
 export UPDATER_OPTS="-DwikibaseHost=${WIKIBASE_HOST} -DwikibaseMaxDaysBack=${WIKIBASE_MAX_DAYS_BACK}"
 
 envsubst < /templates/mwservices.json > /wdqs/mwservices.json