API consumption use cases

[jmatsushita]

Who will be consuming the raw metrics data that OII will aggregate from its measurement partners? The end goal is to help analysts experiment and come up with interesting derived metrics and scores, and specifically scores that can be used by end users to make choices about tools, but there are other potential users.

This needs more work with personas but here are some thoughts:

Project Owners

• Dashboard : interested in getting metrics update in real time for a particular project. Typically 1/ search for project 2/ get all metrics available 3/ get data streams. (A la newrelic)
• Project update : interested in taking ownership of a project ("this is mine!") and adding missing data about it. 1/ start a verification of ownership process (OII might want to trust measurement partner's own verification) 2/ submit proposed changes (some project owner changes should be treated as "identified" reviews, others might be the owner's prerogative to add/change/correct/remove)

  Measurement Partners

Use OII data to enrich existing metrics about a project. (e.g. Libraries.io displaying RDR ranking). 1/ adopt project unique identifier 2/ embed a specific set of metrics (on each page display? cached where?)

Analysts

Want to combine, weight, do complex queries, access raw data to create derived metrics or scores for a range of specific needs (end user security in specific adversary capability contexts, correlation, prediction,...). For this use case the first need that OII can support is as a machine readable data catalogue of available metrics for specific projects.

The analytical needs are likely to be hugely diverse and varied and might end up trying to replicate efforts such as BOA or general purpose data analytics platform.

It might be better to facilitate ingestion of the catalogue metrics into a range of analytical platforms, either by:

• working with measurement partners to make their data easier to upload, or have them continuously feed into real time analytics platforms,
• or it might be preferable to manage "message passing adapters" or "ETL agents" that would do this with a trusted data analytic infrastructure provider.

Partnering with MLab @collina @meredithmeredith to have access to Big Query, Cloud Dataflow might be an option. Or continue working with @hellais and @sachavg on setting up a civil society data analytics infrastructure.

Probably also in the horizon, the model of distributed indexing to enable a federated query interface like Linked Data Fragments is probably relevant. Wondering what @elf-pavlik thinks.

[elf-pavlik]

:+1: http://linkeddatafragments.org/ and hypermedia APIs in general :link: https://twitter.com/elfpavlik/status/635835566328807424

[hellais]

I would image at the very least the ideas implemented for the OONI pipeline, could be adapter to your use case. Perhaps even some of the code could be re-cycled and adapted. If there are a lot of people interested in this and there are plans to invest a serious amount of development time from !OONI people, then some redesigning and refactoring could be done to reduce even more the duplication of code and maximise re-use.