V2

[imulab]

v1 of go-scim was my first attempt to approach an open source project and it was a great experience. However, to be honest, I found myself not entirely up to the task of maintaining this version as it has got several problems:

• Reflection: extensive use of reflection makes the code harder to maintain
• Resource model: the resource model was decoupled with the schema attributes, making it awkward when having to walk the tree with the assistance of attributes. This affects all further operations (i.e. validation, modification) on the resource.
• Filter/Path: filter and path parsing could be improved to be more dependable; the URN prefix feature is somewhat hardcoded.
• JSON: JSON serialization is based on reflection while we already have the type information available in the schema; JSON deserialization relies on an intermediate data structure.
• BSON: Like JSON, serialization to MongoDB also relies on the intermediate data type of bson.M, which is essentially a map. Then MongoDB picks it up and encode it to bytes using reflection. Again, it is wasteful to do the same thing twice.
• SPI: I like the idea of using SPI and that this library serves as a foundation of server implementations rather than being an implementation itself. However, not providing a not-a-toy initial setup is rather inconvenient for people to test this software and customize. I would like to keep SPIs as extension points, but do provide workable defaults out of the box for v2.

Comments and ideas are welcomed.

[requaos]

@imulab In order to provide a working oobe, we will need to supply an example set of configuration files to accompany the Dockerfile and docker-compose.yml files for local testing and development.

[requaos]

@imulab Can we put those files in /server/public/{schemas|sp_config|resource_types} to follow the pattern set in the mongo folder?

[imulab]

V2 has been released, closing this one and take a break.