Evaluate Trellis as LDP Component for Sinopia Server

[cmharlow]

https://github.com/trellis-ldp

[jmartin-sul]

had a short meeting w/ @cmh2166 and @jermnelson to discuss figuring out whether trellis fits our needs. questions christina has:

• [x] state of the codebase, dev practices: adequately tested, well documented, well laid out, well maintained, in active use/development?
  • well maintained, well tested, well decomposed, pretty well documented. i had some questions, but aaron provided some great answers with pointers to specific parts of the code. the postgres implementation of trellis-ext-db was done in part to suss out whether the Java LDP API trellis provides is actually portable between different backend stores. i can post my Q&A with aaron in a separate comment.
• [x] what does deployment mean? expects a lot of infrastructure? production ready release expects maybe a CentOS env running a suite of services (kafka topic, RDBMS or triplestore, etc)?
  • for trellis-ext-db, you just need the trellis-ext-db java server, and an RDBMS like postgres. kafka, JMS, etc are just options for message consumers (to take linked data notifications). kafka etc are not strictly necessary.
  • for plain trellis, a triplestore is needed (or a local directory for it to use its internal Jena/TDB triplestore). plain trellis has the same message consumer optional dependency as trellis-ext-db.
• [x] should have deployment patterns/scripts to spin up. how reasonable are they, how reasonable is AWS deployment? would be great to keep it in AWS, though we might be willing to do on prem if needed.
  • it seems like we could deploy trellis-ext-db instances as AMIs to fargate, talking to an amazon postgres-looking RDS instance.
  • maybe we could write something to emit SNS or SQS notifications (instead of JMS or kafka)? that seems saner than consuming JMS messages and turning them into SNS or SQS.
• [x] dockerized versions of dependencies from the trellis people. how usable are they at present?
  • there are some docker images of trellis and trellis-ext-db, but we might be better off using them as a guideline to config our own docker images.
• [x] how easy is it to add functionality, how extensible is it?
  • seems very extensible. trellis-ext-db implementation speaks to that.
  • [x] what is it missing that we need, what does it have that we don't need?
  • still figuring this out. seems like we probably need:
    • identifier minting
    • passing in a graph with multiple resources in one request to trellis for CRUD
    • ???
• [x] comes with a "read-only triplestore" as a "sidecar", implemented on camel?
  • this is one of the projects in the https://github.com/trellis-ldp/camel-ldp-recipes repo
• [x] last thing to investigate: they have an AWS version in progress, what's the state of it?
  • totally WIP experimental, but we can probably do without it.

@cmh2166, correct me if i got any of that wrong.

[cmharlow]

Done, see: https://github.com/LD4P/sinopia_server/wiki/Draft-Notes-for-Sinopia-Server-API-Spec

[acoburn]

@cmh2166 just an FYI, since I see that you're doing a lot of work with JSON-LD. Trellis supports custom JSON-LD profiles/contexts in responses. So, for instance, if you want a compact JSON-LD response using the Web Annotation @context, here's your request:

curl -i http://localhost:8080/resource -H"Accept: application/ld+json; profile=\"http://www.w3.org/ns/anno.jsonld\""

Because this is (potentially) a security issue -- letting clients download arbitrary web resources -- there is a whitelist in the configuration under

jsonld:
  contextWhitelist:
    - http://www.w3.org/ns/anno.jsonld

You can add as many contexts as you like, or even open up an entire domain as "trusted". These remote contexts also get cached internally to help speed things up.

[cmharlow]

@acoburn thanks for this! i'll try to break up our questions & such so we can get them here and ping you (+ the trellis google groups) more

[acoburn]

@cmh2166 that sounds great! I noticed you had a question about provenance data (who changes what resources and when) -- each resource has a full audit log that you can view with Prefer headers. Documentation is here: https://github.com/trellis-ldp/trellis/wiki/Resource-Provenance--(Audit)

The audit trail doesn't tell you exactly which triples where changed, but you can establish that by combining the data with the Memento representations (if you wanted to know which particular triples were modified by a certain user at a certain point in time).