papers/lp-200-zap-stack/sections/01-introduction.tex

\section{Introduction}

This paper expands on LP-200 (\texttt{lps/LP-200-zap-stack.md}) with
detail on the ten-layer model, the decomplected concerns, the measured
per-operation cost wins, and the explicit composition table tying the
umbrella to its sibling LPs.

The Lux ZAP Stack is the architectural commitment of the new final Lux
network. It is activated at unix timestamp \texttt{1766708400}
(2025-12-25 16:20 Pacific) -- the genesis of the new network. There is
no migration path because there is nothing to migrate: the pre-Quasar
Edition Lux network (2020--2025) is a separate network operating a
different stack, and is out of scope.

\subsection{What was wrong before}

The legacy blockchain stack that the new network replaces -- a
direct descendant of the Avalanche-derived codebase Lux forked from in
2019 -- braided wire, in-memory, and on-disk concerns through a
runtime \texttt{codec.Manager} registry. Every transaction was
unmarshalled into a Go struct on receive, re-marshalled into a
storage row on insert, unmarshalled again from storage on read,
re-marshalled into the gossip envelope on broadcast, and re-marshalled
yet again into the snapshot stream on backup. Five encode/decode
hops per transaction, each paying a reflection cost, each requiring a
runtime codec-version registry lookup, each allocating fresh Go
structs that the garbage collector eventually had to reclaim. The
braiding was load-bearing: removing one hop required touching the
other four.

\subsection{What this LP solves}

The ZAP Stack unbraids the wire from the in-memory representation
from the on-disk record. A transaction is a binary buffer when it
crosses the wire (LP-022); it is the same buffer in the mempool; it
is the same buffer when a block frame references it by offset; it is
the same buffer when ZapDB streams it to an age-encrypted snapshot.
One byte stream, ten orthogonal layers reading and producing bytes,
no codec registry.

The reason this matters is operational, not aesthetic. The cost of
every braided hop in the legacy stack is paid \emph{every block}.
Removing the braids removes the cost. The architectural commitment is
the load-bearing one; the measured wins (Section~\ref{sec:performance})
confirm it but are not the point. The point is that each of the ten
layers is independently complete and orthogonal to its siblings, so
that future work (rollups, light clients, state commitments) composes
without re-litigating the encode/decode boundary.

\subsection{Values, not places}

Rich Hickey distinguishes places (slots that hold different things
over time) from values (things that do not change). ZAP-native is
values all the way down: the buffer that arrives over the wire is the
same buffer in the mempool, the same buffer in the block, the same
buffer in state, the same buffer in the snapshot. The identity is a
function of the buffer; the transcript is a function of the buffer;
the signature is over a window of the buffer. Nothing mutates.
Nothing re-encodes. The buffer is the value.

This is the same discipline applied throughout: profile gating is one
function in one place (LP-217); the wallet builder's fee-asset is one
field with one canonical name (\texttt{UTXOAssetID}); the cert is one
struct shape across all modes (LP-182). One way to do each thing.