Thread
#getting-started
    n

    nutritious-bird-77396

    1 year ago
    @microscopic-receptionist-23548 Could you help me understand the reasoning behind splitting GMA into a separate repo?
    m

    mammoth-bear-12532

    1 year ago
    Hi @nutritious-bird-77396: this was mainly done to decouple the libraries that support storage of metadata from the "datahub application" that contains the opinionated models (what is a dataset) along with services and frontend that serve that model.
    e.g. LinkedIn has its own models internally and its own gms implementations, but uses gma library for storing and retrieving metadata
    Is there a specific problem you see with the split repo approach?
    m

    microscopic-receptionist-23548

    1 year ago
    It also helps us make breaking changes in GMA if needed; we don't immediately need to update dartahub in the same PR
    b

    big-carpet-38439

    1 year ago
    These two things are quite tightly coupled though at least in OSS world. Makes changes spanning multiple areas of DH more difficult with them split
    n

    nutritious-bird-77396

    1 year ago
    ^^ This was the concern for my question @mammoth-bear-12532 For any change that impacts GMA in datahub i have to make changes in 2 repos. Just wanted to know the background.
    m

    microscopic-receptionist-23548

    1 year ago
    Two PRs aren't a bad thing; it also helps to enforce smaller PRs, which are also not a bad thing 🙂
    b

    big-carpet-38439

    1 year ago
    there's always a tradeoff: more PRs means higher friction, means a productivity hit