TAG Meeting 2017-05-03
Pre-read Materials
Ed-Fi Enumerations – Discussion Document.docx
Ed-Fi ODS API Validation Recommendations.docx
Participants
- Dan Retzlaff
- Dirk Bradley (Michigan Data Hub)
- Don Dailey
- Fuat Aki
- Jennifer Downey (Infinite Campus)
- John Raub (WI)
- Jon Berry
- Mark Reichert
- Matt Warden
- Sherod Keen
- Silvia Brunet-Jones
- Tim Casey
- Eric Jansson
- Sayee Srinivasan (Ed-Fi Alliance)
Notes
Enumerations Discussion
See pre-read materials.
One concept discussed was: types are for longitudinal analysis. Can some types be resolved at a category level? Can we develop and govern long-term meaning for these categories?
Vendor viewpoint was raised at several points:
- Types should not be modified
- Short description gets reported for types, why do types have a short description and a code value?
LeaveEventCategoryType was used as example:
- Some descriptions missing - need to add some
- Don't want to use 'other' - how do we do this?
Some agreement that more governance of types is needed: some missing, need feedback on which might be problematic.
API Validation Discussion
See pre-read materials.
Some discussion of history of bulk exchanges: exceptions table in 1.2 Ed-Fi bulk exchanges was helpful. Lots of data ended up in the exceptions table, so the ODS was offering some default support for data quality reporting.
Some participants reported being torn on the question, noting it is far easier to address data quality up front before the data is accepted, but also acknowledging that thsi could result in non-standard ecosystem experiences. One challenge is there are many vendors, and it takes substantial time to work through data quality issues downstream. Core vendor problem was represented as the need to code for different specs. Need some mechanism for vendors to not worry about a certain set of core data.
One validation that has appeared is to block data provided by SIS systems that is not collected by a state. Some states are not allowed to collect data that is not part of a federal reporting - interpreted as the data should not even be sent, not just that the data should not be persisted.
Materials