admin
a7f897a7dd
Read-only. fetch.py + tables.py (stdlib), TABLE.md (45 charts, % of respondents), FELHOM.md (apps, features, reasons, hardware, trends, ten ranked recommendations), METHOD.md (hashes, counts, matching, MIT licence, limits). Survey files are not committed. One register row (R-989) and one STATUS line ask the operator to rule on the recommendations.
3.1 KiB
Method — selfh.st survey read, 2026-10-12
Fetched 2026-10-12 by fetch.py to ~/research-cache/selfhst-survey/ (outside every repo; not committed).
| Year | Data file | Bytes | SHA-256 | Matches the spec |
|---|---|---|---|---|
| 2023 | https://raw.githubusercontent.com/selfhst/cdn/main/assets/surveys/annual/2023-results.json |
10362 | 714350e5e4b8f3ccbc7ce8c1348122a9b7a4ae8167abaf2c051d3dcc9911f03f |
yes |
| 2024 | …/2024-results.json |
14290 | 289c5161e5b6b5684ac6c751c16e73dd0dc42490c5c0d24517a6c1f891da4832 |
yes |
| 2025 | …/2025-results.json |
20428 | e083094ad7f8cb6721f7059c6c73e8779273717bb3d4c1f14dbd744dc0d0229e |
yes |
Respondents. 2025: 4,081, stated as completed responses on https://selfh.st/survey/2025-results/.
2023 and 2024 pages give only rounded numbers ("just shy of 1,900", "~3,700" — https://selfh.st/survey/2023-results/,
…/2024-results/). The exact figures used, 1,869 and 3,655, are the column sum of the first, mandatory,
single-choice question (personal / enterprise / both). The same method gives exactly 4,081 for 2025.
Shapes. 2023/2024: top-level charts, each with an id. 2025: categories[].charts[], no id, five
categories, 39 charts. Some 2025 charts carry a commentary field (not copied).
Percentages. Multi-select charts: % of the year's respondents. Single-choice charts (marked in TABLE.md):
% of the column sum, because the question went to a subset or not everyone answered. Favourite software is one
answer per person but the chart drops small counts, so it is % of respondents.
Matching across years. 2023/2024 by id (same titles). 2025 by question meaning, a hand-written map in
tables.py (MATCH_2025); both titles are printed in every section. Answer labels mapped by hand where the
wording changed (ALIAS, e.g. Yes → Deploy, Generic Linux → Linux, Docker - Individual → Container (Separate),
2023 Paperless → Paperless-ngx, 2024 "Trilum" → Trilium); labels that differ only in case are merged and listed.
Unmatched (no same question in 2025): 2023 offsite-backup yes/no (T40), 2023 Kubernetes (T41 — 2024/2025 ask about any orchestration), 2023 media-streaming yes/no (T42), 2024 smart-home protocols (T43), 2024 ActivityPub yes/no and platforms (T44, T45). 2025-only: Management Platforms, Authentication, Git.
Licence. The results page shows no licence. The data repository github.com/selfhst/cdn is MIT
(GitHub API license.spdx_id = MIT; LICENSE reads "MIT License, Copyright (c) 2025 selfh.st"). We commit no
survey file and no survey prose; only our own counts derived from it, each cited to its URL.
Limits. Hobby users, mostly technical (Technology career 80.6 %, T36), self-selected through the selfh.st
newsletter and community, not households buying a service. Multi-select counts: a column passes 100 %. Open-text
charts (favourites) omit small answers. Country: the JSON files carry no country data, but the 2024 results
page has a country table (HTML only, not in tables.py): United States 964, Germany 430 … Hungary 19.
2023 and 2025 pages have none.