Goals for the next release of neurobank are to address a number of friction points:

1. Lack of meaningful testing. Testing can only be done with an active registry set up, which is difficult to incorporate into a CI workflow. We need a mock registry.
2. Slow requests. As we incorporate neurobank into analyses, scripts are often having to look up as many as 100 resources and/or fetch the resources. These requests all happen serially, with no sharing of HTTP sessions, and can produce major slowdowns. Client scripts often bypass this package entirely.
3. Large and inconsistent API. Many of the most commonly used functions are in the core module, but users often have to dig into the registry module to perform certain tasks. The API is not consistent between the two modules which leads to confusion and error.
4. Lots of duplicated code. The registry module has many functions that are all basically wrappers around calls to requests with specific URLs.