| name | extra-languages-core |
| description | Shared reference for the extra-languages cluster: the one decision every niche stack turns on — write the platform's native, current idiom and contain any foreign/legacy runtime behind an adapter boundary — plus version-pinning conventions, a stack × layer matrix, and the trust-boundary guardrails its spokes share. USE WHEN choosing a niche-stack idiom or default dependency, deciding where a legacy/compat layer goes, pinning a toolchain, or scoping whether work belongs here versus a major-language cluster. |
| cluster | extra-languages |
| version | 1.0.0 |
Extra Languages Core
Shared model for the extra-languages cluster — the deferred-tier home for niche, platform-
specific language and runtime specialists. The spokes are independent stacks, so the value here
is the one cross-cutting discipline they all obey and the rules for keeping the cluster from
drifting into a junk drawer.
1. The decision everything turns on: native idiom + contained boundary
A niche stack fails not on algorithms but on fighting the platform. The rule:
Write the platform's current, native idiom. When you must touch a foreign or legacy runtime,
isolate it behind an adapter — never let its types leak across the app.
This has two halves:
- Native + current. Use the stack's present-day blessed toolkit, not the one your muscle
memory or last year's tutorial reaches for. For Solana that means
@solana/kit first for
client/RPC/transaction code and framework-kit (@solana/client + @solana/react-hooks)
for UI — not web3.js by default.
- Contained boundary. When a dependency demands the legacy runtime (a lib that wants
PublicKey/Transaction/Connection), introduce a single adapter module
(@solana/web3-compat) and keep those types out of the rest of the codebase. The boundary is
a named seam, not a diffusion.
app code (native idiom) ──> [adapter / compat seam] ──> legacy or foreign runtime
^ types stop here ^
2. Scope: what belongs in this cluster
This cluster owns the long tail — stacks with no dedicated cluster of their own. Before
adding or routing, check it isn't already owned:
| If the work is… | It belongs in… |
|---|
| Solana program / dApp / wallet / tx pipeline | here → solana-dev |
| General Rust (no chain) | rust / systems-languages |
| Python / JVM backend | python-backend / jvm |
| EVM / Ethereum / Solidity | blockchain-web3 |
| A niche stack with no cluster yet | here (as a new spoke) |
Graduation rule: when a spoke accumulates enough depth to justify its own orchestrator +
core, promote it out of extra-languages rather than letting this cluster bloat.
3. Stack × layer matrix (current spokes)
| Stack | UI / client | Core logic | Toolchain & tests | Spoke |
|---|
| Solana | framework-kit (@solana/client + react-hooks); Wallet Standard connect | @solana/kit types/codecs; @solana-program/* builders; Anchor (iterate) or Pinocchio (CU/footprint) programs | Anchor/Solana CLI pinned; LiteSVM/Mollusk (unit), Surfpool (integration); Codama codegen from IDL | solana-dev |
As spokes are added, extend this table — one row per stack, same three layers, so the
orchestrator can route by (stack, layer) without reading each spoke.
4. Version & toolchain conventions
The long tail breaks on version drift more than on logic. Therefore:
- Pin and state the language/CLI/framework version for any work (e.g. Anchor + Solana CLI
pair). Solana's moving target lives in
solana-dev's compatibility-matrix.md.
- Recognize toolchain failures as toolchain failures — GLIBC errors, ABI/CLI mismatches, and
dependency conflicts are version problems, not code bugs; route them to the spoke's
common-errors.md, don't rewrite working logic.
- Prefer the newest blessed API surface; quarantine legacy in the adapter seam (§1).
5. Shared guardrails
- Native idiom first; legacy/foreign runtime only behind a named adapter boundary.
- State every trust-boundary fact before acting — for on-chain work: cluster + RPC/ws
endpoints, fee payer + recent blockhash, compute budget, expected account owners/signers/
writability, and the token-program variant (SPL Token vs Token-2022 + extensions).
- Never silently widen a security boundary (signing authority, CPI target, account
writability, scope of a compat adapter) — call it out.
- Pin the toolchain; treat version mismatches as the first hypothesis for niche-stack breakage.
- Deliverables: exact files changed + diffs, install/build/test commands, and a short risk
note for anything touching signing/fees/CPIs/transfers.
- Keep the cluster honest: only long-tail stacks live here; graduate any that outgrow it.
