As I'm doing anyway the work required for a saner Eulora client, I'm stuck as well with writing a minimal documentation for it, since on one hand there's no one else who can quite write it anyway and on the other hand there is no such thing at all¹ for the old piece of inherited gnarl so the slate is blank, the space infinite and the need impossible to know in advance. So I'll just jot it down here for future reference and if you have anything to say about it, the sooner you write it down in the comments section below the better for you.

As general structure, the current plan for the new client is to have essentially 2 main parts: one is the actual Core, written in Ada and - hopefully² - provided as a standalone library, blissfully unaware of ANY of the C/CPP legacy tangle; the other is an updated version of the current C/CPP tangle of a "client" that will use the Core for pretty much everything except the actual GUI. The GUI remains provided via CrystalSpace and cal3d, as it currently is. In other words, the Core offers an implementation of a basic Euloran client, to be used from/by any UI, graphical or otherwise as you care to make for yourself. Note that the interface is NOT the only thing you can build on top of the client Core. Essentially any client logic, bots and anything of the kind would still be the sort of thing built on top of / around the Core. A quick diagram of the whole thing looks like this:

The Core itself is currently made of 3 conceptual layers, in order from bottom to top: encryption utilities (EuCrypt), SMG communications (SMG_Comms), client data cache. The encryption utilities are those of EuCrypt: CRC32 calculation, Serpent encryption/decryption, Keccak encryption/decryption, RNG interface using Fuckgoats, RSA key generation, RSA encryption/decryption using TMSR OAEP schema. The SMG communication part provides mainly: UDP with fixed-size but configurable payload, in/out message queues and attached "consumer" tasks, config file reader, packing/unpacking (i.e. raw, bottom-most layer of SMG communication protocol), read/write of messages conforming to SMG communication protocol specification, thin layer on top for encapsulating a "communication link" i.e. required addresses, ports and keys for both ends. The client data cache keeps effectively the whole world of Eulora as the client knows it at some specific time and provides thread-safe getters and setters for anything and everything that it can possibly hold.

Note that the client's knowledge of the world is by definition a cache as it flows always from the server and entirely in response to client's own requests - there is by design no "push" of information or anything else from server to client at any time. So the client may request from the server anything and everything it wants to³, having however to decide for itself what to make of the answers it receives and even what "obsolete" means in various contexts. At the moment this cache is still in the works and for this reason the least fully-fledged but at any rate, it has to grow around the structuring of data that is discussed in the SMG protocol specification: a potentially infinite hierarchical structure of "objects"⁴ having each a subset of an overall set of potential "properties"⁵ with their corresponding "values"⁶. Due to this trinity of objects-properties-values with its corresponding trinity of ID spaces, I took to calling the whole thing a tricache but the name as the design are still in the works so they'll have to enjoy their own detailed description at a later date. Until then, what have *you* been working on and how is that going?

1. Perhaps you count this spaghetti piece as documentation but even so, just note who extracted it in the first place, anyway. ↩

2. I'm holding quite tightly to this idea as I certainly do NOT want to have it depend on any of the current C/CPP mess. This being said, I can't know in advance what I might still uncover so for now and until it's all done, the qualifier remains "hopefully" rather than surely. ↩

3. This in practice is of course quite severely restricted by what it *knows*, unsurprisingly. ↩

4. Where "world", "map", "actor" or "pot" are equally objects and not even the only objects. Essentially an object is anything that is present in the game's world. ↩

5. A property is a pre-defined and public structure that specifies the *type* of values that make up the property and their actual meaning. ↩

6. The values may be given directly (e.g. "3" or "somename") or by reference aka as an ID. This ID may be itself either the ID of a suitable object or the ID of a value that is essentially too big to pass around every time - in the protocol's specification this sort of ID is currently called for historical reasons "filename" but it's worth noting that there is no imposition for it to be stored as a file at any point. ↩