Dev notes from c4l15 (and beyond)
- Anna Headley (Unlicensed)
Owned by Anna Headley (Unlicensed)
Current priorities in Hydra community
- hydra works application profile for data modeling
- https://wiki.duraspace.org/display/FF/Fedora+Community+Data+Model
- note: namespacing can be changed at any time.
- predicates used can be changed at any time.
- Ordering model is part of this work. Tricky and a collaborative solution would be useful.
- implications of modeling on LDP (linked data platform, an API)
- Other types of metadata: descriptive, rights (access permissions, not in the legal sense), technical, relationships (focus up to now)
- web ACL (this is used by both hydra and islandora but could be shared?) - web access control layer (in hydra this is an RDF ontology)
- audit service. fedora is putting together some feature sprints for march into april. requirements gathering is still to come. sounds like reporting? audit service to collect info about what’s happening in your repo. premise service collecting it. analytics service also might sit on top of an audit service. change events - what happened and who did it? checksums and checksum checking is a part of this.
Gleaned info
- Fedora 4 stores objects in RDF datastreams. Ordering is difficult in a graph. There are multiple ordering systems. ordering with properties vs. ore proxy (blank node that links to ordered thing, set, and doubly-linked list to prev / next? Proxy solution is the current proposal.
- If you have a ton of different collections with different types of data, what about context-sensitive facets? This is done programmatically
- What are the options for authentication? Talk to Kate Lynch; they are using native-only (database authentication)
On cloud hosting
- Justin says they do AWS all the time.
- Most resource expensive part of the system is derivatives generation.
- NFS - SAN alternative? On-site with tape backup. talk to the folks at penn state.
- back end and front-end should be in the same data center for speed of access.
- Glacier storage (but just to dump and never access)
- amazon does not have an acceptable SLA for preservation.
- security could be an issue as well with amazon (concern for Hathi trust because they have copyrighted material)
- DPN - http://www.dpn.org/about/
- http://www.evault.com/
- https://www.digitalocean.com/
- hydra wiki hardware page - https://wiki.duraspace.org/display/hydra/Deployment+Hardware+Information
Will Sufia work for us?
- How important is the self-deposit aspect of Sufia? Can Sufia be leveraged into a more traditional resource storage / metadata creation tool?
- Deposit can be locked down by group / role.
- What about worthwhile?
- Sufia has only one data model (file). No such thing as a compound object. Worthwhile has more than one data model. However, more are expected to be added to Sufia.
- worthwhile also is not on fedora 4. but this may happen.
- case.curationexperts.com (example of worthwhile - non-persistent)
- Posted to the list by mjgiarlo:
* When do you need multi-file works? If you need them NOW, Worthwhile is probably a better fit. If you can wait some months, Sufia will support them (and will implement them in a way that is faithful to the emerging Fedora Community Data Model).
* When do you need Fedora 4 support? If you need them NOW, Sufia is probably a better fit.
If you need both now... you've got a nice chunk of work for yourself in the coming months. ;) (By which time we may have an initial implementation of multi-file works under way in Sufia.)
- examples:
- github/KevinSmithLibrary/absolute - application using worthwhile
- sufia: scholarsphere at penn state