Research
This folder keeps the research that Apollo Download Manager's decisions were based on, so a decision can always be traced to its evidence. Research documents are inputs, not decisions. When a decision overrides a research recommendation, the decision record says so. Decision records live in docs/decisions/.
| Date | Document | Status | Notes |
|---|---|---|---|
| 2026-10-21 | Velocity libcurl download engine — research report, revision 2 | Research baseline for Apollo DM | Written about V1 at e9d717a8. Sections 4 (architecture) and 12 (recommendation) shaped ADR-0001. Its aria2 assumptions (legacy engine, fallback, migration of aria2 downloads) do not apply to Apollo DM. |
| 2026-10-22 | Engine prototype results | Evidence for #38 | Apollo DM's own findings from the #37 prototype: correctness in both modes on Windows and Linux. Performance not measured. |
| 2026-10-22 | Engine mode costs | Evidence for #38 | What the child process costs against in-process: throughput, CPU, memory, latency, startup, shutdown and crash detection, on Windows and Linux (loopback). |
| 2026-11-05 | Product name shortlist | Input for #346 | Six candidate "<Name> Download Manager" names, checked for existing download managers, other software and trademark aggregator records. The official trademark databases were not searched; Singlbase does the legal clearance. |
| 2026-11-06 | Torrent peer discovery, trackers and performance | Input for #352 | What the torrent engine sets and leaves to libtorrent, a measured baseline on Windows, the public tracker options for the PO, and the follow-ups in order |
Copies of outside research here are verbatim exports. Corrections go into a new revision with its own date, never into an old copy.