Mathhub.info needs to know what all the gropus and archives are (visible for a given user) on http://gl.mathhub.info.
We need a service that collects information about the groups/projects from the GitLab server, reads all the <group>/meta-inf/GROUP_MANIFEST.MF and <archive>/META-INF/MANIFEST.MF files and aggregates the relevant metadata from these into a database which it makes available to MMT. In particular, we need the MMT archive tags.
Then MathHub can query this service (from time to time) to find new archives (of a given tag; e.g. OAF or ODK; see #25 or #26, oaf.mathhub.info should be the MathHub with the OAF-tagged archives) and load them into the MathHub MMT server. Another thing the service should know about are archive-level dependencies, so that MathHub.info can always have a self-contained collection of archives.
@tkw1536 thinks this servis should be separate server that talks REST; I frankly do not care. But we need something like this.
NOTE that what is visible on gl.mathhub.info depends on the logged-in user, for instance MKohlhase has a lot of non-public archives e.g. in the MiKoProblems (problems with solutions) and MiKoCourses (exams and homeworks) groups. Eventually, the inventory should be able to impersonate MKohlhase and give personalized answers in this situation as well. This will probably mean passing around private keys. But for the moment we should concentrate on the public inventory.
Mathhub.info needs to know what all the gropus and archives are (visible for a given user) on http://gl.mathhub.info.
We need a service that collects information about the groups/projects from the GitLab server, reads all the
<group>/meta-inf/GROUP_MANIFEST.MF
and<archive>/META-INF/MANIFEST.MF
files and aggregates the relevant metadata from these into a database which it makes available to MMT. In particular, we need the MMT archive tags.Then MathHub can query this service (from time to time) to find new archives (of a given tag; e.g. OAF or ODK; see #25 or #26, oaf.mathhub.info should be the MathHub with the OAF-tagged archives) and load them into the MathHub MMT server. Another thing the service should know about are archive-level dependencies, so that MathHub.info can always have a self-contained collection of archives.
@tkw1536 thinks this servis should be separate server that talks REST; I frankly do not care. But we need something like this.
NOTE that what is visible on gl.mathhub.info depends on the logged-in user, for instance MKohlhase has a lot of non-public archives e.g. in the
MiKoProblems
(problems with solutions) andMiKoCourses
(exams and homeworks) groups. Eventually, the inventory should be able to impersonate MKohlhase and give personalized answers in this situation as well. This will probably mean passing around private keys. But for the moment we should concentrate on the public inventory.