API endpoints insufficient for mirroring

So I've started to more seriously toy with the idea of using one of my servers as a true, publicly-accessible third party mirror for Keybase. Unfortunately, it looks like the API as it currently stands does not provide sufficient endpoints for achieving this goal. Or rather, I should say that it doesn't provide sufficient endpoints for achieving this goal with any degree of network efficiency.

The core problem is that there is no (good) way to enumerate the set of all users. It's technically possible to trick the API into this functionality via the /autocomplete endpoint, but doing so would require trial-and-error enumeration of various alphabetical prefixes, each of which entailing an HTTP round-trip. Terrifying to say the least.

What I would rather have is some way to stream the directory of all user JSON objects from Keybase to my server in a relatively direct fashion. Presumably, this would be a very large amount of data (hence the streaming), and requests to this endpoint could and should be throttled significantly. As an additional safety measure, you could attach an API key validation to this special endpoint, making it slightly easier to block abusers.

In general though, I'm pretty sure this is a requisite first step to getting real read-only mirroring in place.

Start by reviewing the existing /autocomplete endpoint and the API surface relevant to user enumeration. Define the streaming directory endpoint, including throttling and API key validation requirements. Done means a documented, network-efficient way to stream all user JSON objects for read-only mirroring.