Architecture Choices

Potentially viable architectures for a cross-chain aware name system.

All somewhat known name systems to date have been restricted to a single chain only which limited their architecture options.

Using general message passing we gain access to a few more design decisions we otherwise would not have.

The big underlying option is to not store all data on a single chain but move data to other chains in some cases for efficiency, security or gas cost reasons.

We will go over both options in detail for you to understand the reasoning behind our choice:

Hub And Spokes Architecture

This architecture uses a single blockchain to store all state data. Changes to that data can be requested from the connected spokes using GMP.

Pros	Cons
Looking up data is straightforward, all queries completed on one blockchain	Writing data from a spoke always requires GMP adding cost and delays
Everything can be done in a single roundtrip between Hub & Spoke chain => 2 messages	Most actions will require the full roundtrip
Battle tested and used in many projects	Not an innovation
	Hard time integrating with other name systems on other blockchains

Simplified Query Algorithm

1. Lookup the name on the hub chain a) Entry is found Entry is your result -> Exit b) No entry is found Name does not exist -> Exit

"Pull Out" Architecture

Instead of the names being stored on just a single hub they can be pulled out onto other chains individually. Meaning that you can store records / ownership of names on other chains.

This will not include sub-names though, those will individually need to be transferred out.

Pros	Cons
Mostly straightforward implementation	Extendability on other chains (e.g. integrating other name systems) always requires GMP
Trading & Record management does not involve GMP	Managing sub-names will require 1-2 GMP messages
Reading records takes at most 2 blockchain lookups

Simplified Query Algorithm

1. Lookup the name on the hub chain a) Entry is found Entry is your result -> Exit b) Link is found Do 2. c) No entry is found Name does not exist -> Exit

2. Lookup name on the chain the link points to. This will always point to a valid entry which is your result.

Distributed Tree Architecture

This architecture mimics DNS to some degree. Data is not stored centrally but only the root node . as well as the tld's are on the Hub chain. The ownership of subdomains can be transferred to other GMP connected chains leaving the Hub chain with nothing but a reference to the other chain.

Pros	Cons
The average user only has to use GMP once for registering and never after	Complicated name hierarchies can result in certain actions requiring an arbitrary high amount of GMP hops
After moving your name you can manage it from any chain without using GMP	Name queries can require reading data from a series of blockchains

Simplified Query Algorithm

Go down the tree level by level. This means you gotta split your name at the . first. (Continue here) Every step you reappend one section to the queried name (right to left) till you are looking up the entire name again.

Looking up a name goes as follows:

1. Check registry on the current chain (Initially the Hub)

2. A: Entry (no link) is found: If the current query is not yet the full name you are looking for than the name does not exist. Otherwise you just found the entry that you were looking for -> Exit B: Link is found: Set current chain to where link points -> Continue C: No entry is found: Name does not exist -> Exit

=> Essentially just follow the links until you either reach the name you are looking for or a dead end