SNMP Research Notes

This file captures actionable takeaways from inspecting other open-source SNMP printer discovery/management tools so we can fold the best ideas into PrintMaster without blindly copying code.

Sources Consulted

• Ircama/epson_print_conf (Python) — exhaustive remote-mode OIDs plus EEPROM decoding for dozens of Epson models (epson_print_conf.py).
• CUPS SNMP backend (backend/snmp.c) — hardened discovery logic with vendor-specific fallbacks (OpenPrinting/cups).

Standard OIDs We Should Poll Everywhere

These came up repeatedly in both projects and map cleanly to our existing data model:

Epson Remote-Mode Command Surface

Epson exposes a remote-control endpoint rooted at 1.3.6.1.4.1.1248.1.2.2.44.1.1.2.1.<cmd bytes…> where the two ASCII bytes identify the command and the next two bytes encode payload length. The Python client simply reuses this helper for several high-value calls we can mirror:

Implementation tips pulled from epson_print_conf.py:

• Read operations use read_key (two-byte secret per model) plus opcodes 65/190/160 before the address pair; writes use opcode 66/189/33 followed by the Caesar-shifted write_key.
• The helper already batches up to three OIDs per PDU via cluster_varbinds; we should copy the idea by grouping EEPROM reads so we do not stall the scanner.
• Several commands (st, rw, vi) return framed ASCII strings containing multiple values; building a lightweight parser (similar to their status_parser) would let us translate Epson-specific alerts into our unified telemetry stream.

Epson EEPROM Windows Worth Mirroring

The configuration table shows consistent address blocks we can codify in agent/scanner/vendor/epson.go once we add EEPROM support:

• Waste/maintenance counters: most EcoTank/XP devices store the first box in addresses 24/25/30, second box in 26/27/34, and flag thresholds at 46/47 (or 54/55 for newer models). Large-format or tri-box units add a third counter at 252/253/254 with threshold 255.
• Cleaning & usage stats: sequences such as [147,149,148] (manual/timer/power cleaning counts) and [171-168] (total print passes) recur across L-series. Rear-feed totals ([755-752]) and scan counters ([1843-1840]) exist on every ET-27xx/28xx/48xx variant.
• Serial/MAC blocks: legacy models keep serial ASCII in range(192,202) while Wi-Fi MACs live at range(130,136) or (newer) range(1920,1926). These ranges match what we already scrape via standard MIBs, so they make excellent fallbacks when the printer restricts host MIB access.
• Reset operations: raw_waste_reset dictionaries show the exact EEPROM values Epson utilities write during a maintenance reset. We should not expose writes in the agent, but understanding the pattern helps us detect when a third-party reset happened (sudden drop to zero combined with unchanged counters).

Discovery and Vendor Fallbacks from CUPS

The CUPS backend reinforces a few best practices we should adopt inside agent/scanner/pipeline.go:

• Always begin with hrDeviceType (1.3.6.1.2.1.25.3.2.1.2) probes to confirm the target reports itself as Printer(3) before issuing heavier walks.
• After the initial response, immediately parallelize GETs for description, IEEE-1284 device ID, location, and URI (ppmPortServiceNameOrURI). This gives enough data to decide whether to keep probing or move on.
• Maintain a table of vendor-specific device-ID OIDs for common manufacturers: e.g., 1.3.6.1.4.1.11.2.3.9.1.1.7.0 (HP), 1.3.6.1.4.1.641.2.1.2.1.3.1 (Lexmark), 1.3.6.1.4.1.367.3.2.1.1.1.11.0 (Ricoh), 1.3.6.1.4.1.128.2.1.3.1.2.0 (Xerox). CUPS hits those opportunistically whenever it sees a response from the matching enterprise OID.
• If the device never returns a URI, CUPS still attempts TCP probes on 9100 (AppSocket) and 515 (LPD) before giving up. We can reuse that idea inside our liveness stage to classify “unknown but listening” devices.

Action Items for PrintMaster

1. Add a vendor plug-in for Epson remote mode: reuse the command table above behind a feature flag, deserialize ST2 payloads, and surface ink/waste metrics in the agent database.
2. Extend the discovery stage with vendor-specific ID OIDs: add a snmpTargets slice similar to CUPS so we can learn make/model even when printers neuter the Printer-MIB tree.
3. Batch EEPROM/SNMP reads: adopt the cluster_varbinds idea so we cap PDUs at three OIDs but still parallelize multiple PDUs; this will keep slow printers from starving an entire worker pool.
4. Persist learned EEPROM ranges: cache which address blocks responded per device so future scans do not brute-force every model-specific range.
5. Map Epson waste counters into our metrics: once we trust the data we can store normalized percentages for main_waste, borderless_waste, and any third maintenance box inside the metrics table with the same downsampling policy as toner levels.