Repository navigation
proposed mdbook structure #19
Copy link
Copy link
Open
Description
- mdbook materials should target 3 main audiences
- users (for introductory, problem definition and overview stuff, not necessarily for all details)
- bitcoin developers
- privacy researchers
- README should be for the repo, addressing potential contributors as well
with that the rough structure i propose for the mdbook:
- introduction (README) - README do-over #74
- problem definition - Define the on-chain privacy problems being addressed #7
- protocol suite
- overview - Add an overview #51
- threat models
- privacy - counterparties, 3rd parties, with or without aux info, ...
- censorship resistance
- liveness
- market mechanism
- DoS / flood protection
- coalition formation
- transaction construction (honest/semi-honest base + bft extensions)
- optimistic track: blocks and phase states - DAG spec #67
- contacts and interactive payments
- recovery track: block dag - DAG spec #67
- validity proof details - validity proof spec #68
- transport layer - transport strawman ladder outline #27
- abstract peer channels (authentication patterns, anonymity)
- abstract broadcast channel
- per session
- open gossip (bft / coalition formation only)
- ownership proofs
- rate limiting
- set reconciliation (naive, rbsr, rbsr+riblt)
- PSBT scrubbing - Add scrubber bip #12
- concrete transports
- peer messages (reference: concrete fields, not quite a wire specification)
- privacy optimization strategy (a recommendation, unlike the protocol suite which is normative)
- optimization framework (abstract: generate decisions and evaluate them, pure vs. mixed strategies) - Abstract cost function #15
- pure strategy evaluation
- cost function structure (a family of cost functions): piecewise linear payoff curves and time dependence, linear combination of terms, bracketing
- privacy liability terms (formulae or pseudocode front loaded)
- graph randomization, anti sybil - add graph randomization / sybil resistance cost function term discussion #22
- sybil resistance index
- others' inputs subset sum density
- critical section vs. ahead of time
- when is approximation valid/useful
- own inputs or outputs subset sum ambiguity
- max flow based counterfactual path estimation
- consolidation penalty
- utxo pool diversity
- ad hoc
- sibling output penalty
- fingerprint avoidance
- graph randomization, anti sybil - add graph randomization / sybil resistance cost function term discussion #22
- tactics (generating the decision tree that gets evaluated)
- bottom up decision tree generation
- coin selection strategies
- change decomposition
- radix denominations - Radix coinjoin denominations notebook #14
- mixed strategy
- randomization (k best sufficiently dissimilar pure strategies)
- time dependence
- internalization
- decision loop (greedy search, depth 1 look ahead)
- full BIP drafts (separately from their discussion in the protocols)
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
No labels
Activity
xstoicunicornx commentedon Aug 5, 2026
Do you think we should try to include these components?
bc1cindy commentedon Aug 5, 2026
looks good! where do #10 and #14 fit in this tree?
zealsham commentedon Aug 5, 2026
i think they would be mentioned in the overview section since they fall under that. Good thing is the mdbook is a living document that would go through rounds and rounds of changes
nothingmuch commentedon Aug 5, 2026
yes but not as standalone sections, coalition formation covers the market based stuff, the cost function and protocol specs should discuss incentive compatibility, privacy preservation is mainly for the cost function, and batching & net settlement are applications of the transaction construction protocols
nothingmuch commentedon Aug 5, 2026
#10 is mainly related to the transports, in some sense it is a way of building a p2p encrypted channel when all there is is a broadcast channel (specifically sneakernet via shared directory or USB sticks), but it's not really a dependency of fungi it's more to do the honest threat model only (since that kind of "transport" can't make any reasonable liveness or anonymous broadcast guarantees), so i think long term it would probably be spun off into its own separate thing (and for all the BIPs eventually they should be upstreamed to the BIPs repo)
#14 would be a subsection in the cost function discussion re subset sum density
[-]proposed structure[/-][+]proposed mdbook structure[/+]