Why Holster
Holster was started because applications built on GunDB kept running into problems. The biggest was that writing a lot of data to disk caused Radisk to stop returning acknowledgements, which led to missing data or even unusable files. A lot of time went into fixing this in the GunDB code, but it wasn't possible without major changes.
Fortunately, the concepts behind GunDB and much of its code were also available in Porting Gun. That code is much more readable, and GunDB's creator Mark Nadal walks through it in a talk at Nordic.js. Holster builds on it, so it's still an implementation of the Gun specification, and aims to have the same features.
Goals
The main goals so far have been readable code and test coverage. Load testing disk writes has also been more important than adding new features, since reliable storage is the reason Holster was started.
Holster is used in production by applications including I Haza Website and RSStream, which helps make sure existing functionality works as expected and shows what needs to be added next.
Holster has also focused on realistic data throughput from both the server and the browser, controlling what is listened to and what is stored. It aims to provide distributed data sync, but within the limits of the devices it is running on. Whether you're only using Holster in the browser or providing a hosted service, you should be able to control who stores data on your devices and the bandwidth allowed to access it.
What's next
Now that Holster is proving reliable for existing applications, it's time to start looking at what other features are possible. WebRTC would be a great addition and can be ported over from GunDB.
There are other great features in GunDB too, such as mesh networking with DAM. Besides porting code, there are plans to add federation like in holster-router, and application namespaces within the distributed graph would be useful too.