agent: ConnectTarget type for connection link or SimpleX name#1796
Open
shumvgolove wants to merge 14 commits into
Open
agent: ConnectTarget type for connection link or SimpleX name#1796shumvgolove wants to merge 14 commits into
shumvgolove wants to merge 14 commits into
Conversation
shumvgolove
force-pushed
the
sh/namespace-agent
branch
from
0b334b6 to
5cf3853
Compare
epoberezkin
reviewed
Jun 4, 2026
Comment on lines
+1610
to
+1612
| nameStart = | ||
| () <$ A.satisfy (\c -> c == '@' || c == '#') | ||
| <|> () <$ "simplex:/name" |
Member
There was a problem hiding this comment.
Suggested change
| nameStart = | |
| () <$ A.satisfy (\c -> c == '@' || c == '#') | |
| <|> () <$ "simplex:/name" | |
| nameStart = "@" <|> "#" <|> "simplex:/name" |
Adds `ConnectTarget = CTLink AConnectionLink | CTName SimplexNameInfo` in Agent/Protocol.hs next to AConnectionLink. The StrEncoding parser gates the CTName branch on a `@`/`#`/`simplex:/name` discriminator so that bare tokens (which SimplexNameInfo accepts for Markdown's sake) cannot ambiguously match at this layer. JSON is a plain string via strToJEncoding, mirroring AConnectionLink. No consumers yet.
Three cosmetic cleanups flagged in code review: - Drop the vestigial () placeholder on decodesSuccessfully; use a prefix call instead of operator style. - Use Data.Either.isLeft instead of a local one-liner. - Add a symmetric CTLink JSON assertion to pin both branches of the wire shape, not just CTName.
Stored as TEXT via decodeLatin1 . strEncode and decoded via fromTextField_ $ eitherToMaybe . strDecode . encodeUtf8 — the established pattern for typed TEXT columns in this codebase (see RcvSwitchStatus, SndSwitchStatus, RatchetSyncState at Agent/Protocol.hs:614-647). Lets simplex-chat carry the type directly in DB tuples without a per-call decode helper.
The instance landed in the previous commit alongside ToField but has zero consumers in either repo — chat-side row decoders use the soft-degradation `decodeSimplexName` helper at the tuple level, never this instance. `fromTextField_` raises ConversionFailed on parse failure, which doesn't compose with the chat policy. Keep ToField (used by parameter binding in name lookups). Leave a comment explaining why FromField is absent so a future contributor doesn't reintroduce it without thinking about the decode policy.
Client.hs: proxyResolveName wraps proxySMPCommand for RSLV/NAME. Agent/Client.hs: resolveName routes via sendOrProxySMPCommand; direct path throws TENoServerAuth since SResolver has no direct client role. Agent.hs: resolveSimplexName takes resolver SMPServer + SimplexNameDomain, looks up the TLD contract (mirrors hardcodedTldRegistries server-side), and forwards the RSLV. TLDWeb is intentionally unmapped.
shumvgolove
force-pushed
the
sh/namespace-agent
branch
from
5cf3853 to
90de60e
Compare
RSLV was previously rejected by the server unless forwarded via PFWD, making the agent's direct fallback unreachable. Relax the server-side guard so RSLV is accepted both forwarded (preferred - hides client IP from the resolver) and direct (faster, exposes client IP). Mode choice is delegated to the client and the operator network config. - Server: drop the forwarded-only check on SResolver in verifyQueueTransmission. - Protocol: give SResolver a client role (SRMessaging) so SMPClient can connect in the Resolver role. checkRole accepts this because RSLV clients have no service binding (falls through to True). - Client: add directResolveName mirroring proxyResolveName via sendProtocolCommand with no auth and no entity (RSLV is noAuthCmd). - Agent: wire the direct path through sendOrProxySMPCommand so the PFWD-or-direct selection works the same as other commands.
Convergence on the Python resolver shape (PR #1795 `snrc-resolve.py`) so a names router can be backed either by the direct-ETH-RPC resolver or by the Python REST resolver without changing the wire format clients see. Wire-level changes: - Add `nickname`, `website`, `location`, `simplex.contact`, `simplex.channel`, `ETH`, `BTC`, `XMR`, `DOT`, `resolver` (SNRC contract address that produced the record); all but `name`, `owner`, `resolver` are optional. - Drop `displayName` (now `name`), `channelLinks`, `contactLinks`, `adminAddress`, `adminEmail`, `expiry`, `isTest`. - The wire NameRecord no longer carries `expiry`; clients trust the server's filter. Expiry checking moves into `decodeGetRecord`, which now takes a `nowSec :: Int64` argument (the placeholder remains, but the field-layout-aware decoder will apply the filter once it lands). - Testnet status is derived from the queried TLD (`TLDTesting` vs `TLDSimplex`) rather than an in-record flag. Other: - ToJSON/FromJSON are hand-rolled because Aeson TH doesn't accommodate dot-keys (`simplex.contact`) or uppercase coin keys (`ETH`/`BTC`...). - `NameLink` newtype is removed (no longer used internally); per-field byte caps are applied directly in the FromJSON parser. - Update the canonical-encoding spec in protocol/simplex-messaging.md.
Adds tests/RSLVTests.hs covering the RSLV pipeline against a running SMP server with a stub ethCall injected via newNamesEnvWith. The production decodeGetRecord is a placeholder (returns Right Nothing for any non- malformed buffer), so the success path is marked pendingWith until the SNRC ABI codec ships; everything else - direct vs PFWD acceptance, contract mismatch, unknown TLD, backend NotFound, transport error mapping, and rslvDisabled - exercises the wiring end-to-end. Adds a minimal test seam: newEnvWithNames / runSMPServerBlockingWithNames that accept a pre-built NamesEnv and skip the real pingEndpoint probe. The production newEnv / runSMPServerBlocking delegate through with namesOverride = Nothing, so behaviour is unchanged outside tests. 7 active tests pass, 1 pending. The existing 42 SMPNamesTests still pass.
The TLD->NameOwner placeholder mapping was duplicated literal-for-literal between Agent.hs (tldNameContract) and Server/Main.hs (hardcodedTldRegistries). Lock-step bumps risked silent divergence. Extract into Simplex.Messaging.SimplexName.Contracts.tldContract; both agent and server read from there. Server-side per-operator TLD config (TldRegistries record, lookupTldAddress, NamesConfig.tldRegistries) is removed entirely - it was already inert post-b66d9730 (which dropped the INI keys).
Replaces the placeholder Right Nothing return with a real ABI decoder for
the assumed Solidity signature:
function getRecord(bytes32 node) external view returns (
string x10 fields, address owner, uint256 expiry
)
Layout: 12 head slots (10 string tail offsets, owner address, uint256
expiry) followed by length-prefixed string data in declaration order.
Server-side expiry filter (nowSec passed by Names.hs:fetch) keeps the wire
NameRecord free of an expiry field. The on-chain value 0 means "never
expires" (reserved names); the caller can pass nowSec = 0 to disable the
filter in tests. nrResolver is populated from the contract address the
server's eth_call was sent to, since the ABI return doesn't carry it.
Zero owner remains the NotFound sentinel.
Drops the placeholder-warn IORef plumbing that surfaced the stub in logs;
the decoder is no longer a stub. Tests that used (32 * 8) sentinel buffers
move to (32 * 12) to match the new head size. Adds an encodeRecordAbi
helper in SMPNamesTests for end-to-end testing; both RSLVTests and the
agent ResolveNameTests reuse it for the success-path tests.
If the SNRC contract ships with a different return layout, this decoder
will need rework; the placeholder gave a documented MVP unblock until
that point.
The hand-rolled NameRecord ToJSON instance produced alphabetical keys (via Aeson's KeyMap canonicalisation), while the hand-rolled toEncoding preserved spec declaration order. The two paths were independent code and could drift on the field set as well, silently breaking the "byte-identical canonical encoding" requirement on the wire path. Top-level TH-splice in Protocol.hs is blocked by dense forward refs between the NameRecord declaration (~line 780) and BrokerMsg's NAME constructor (~line 880). Extract NameRecord and NameOwner into two new self-contained modules that support TH cleanly, alongside the existing Simplex.Messaging.SimplexName tree. NameRecord's ToJSON is now TH-derived with a custom fieldLabelModifier covering dot-keys (simplex.contact / simplex.channel) and uppercase coin keys (ETH/BTC/XMR/DOT). omitNothingFields is set to False to preserve the previous wire shape (absent optionals emitted as JSON `null`). FromJSON stays hand-rolled to enforce per-field UTF-8 byte-length caps that TH cannot express. Note: the canonical Value-encoding path (J.encode . J.toJSON) still re-emits keys alphabetically because Aeson's KeyMap is internally sorted; only the wire path (J.encode -> toEncoding) preserves the spec order, and only that path is part of the protocol contract. The new "toJSON and toEncoding agree on the field set" test pins the no-drift-on-field-set invariant for future edits. Protocol.hs re-exports both types so downstream callers are unchanged.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.