Commit e669d57
docs: refresh the comparison table for the 2.4.0 feature set (#180)
Six new rows in "How Driven compares" for what landed since v2.3.0, plus
a
re-verification of every competitor cell I touched against current
upstream
docs. Docs-only; no code or `site-landing/` changes (the landing page
has no
mirror of this table).
## Rows added
| Row | Why |
| --- | --- |
| Re-uploads backup copies deleted at the destination | #171 (audit) +
#168 (live self-heal) |
| Parallel, multi-threaded local scan | #169 |
| OS-level CPU / disk I/O priority for backup work | #170, #173, #176,
#179 |
| Live preview of which files a rule keeps or drops | #172, #177 |
| Rolling local logs plus a one-click diagnostics bundle | #167 |
| Reproducible end-to-end benchmark suite in the repo | #178 |
Also: a `:grey_question:` legend entry ("not documented"), so a
closed-source
client whose behaviour Google or Backblaze simply does not publish is
marked
honestly instead of being guessed at; the intro paragraph now mentions
the
priority work; the Features list gains the parallel scanner, the
preview, the
priority setting, the remote audit, and rolling logs; and `just bench`
is in
the recipe list with a pointer to `bench/README.md`.
## On the benchmark numbers
I deliberately quoted **no** upload throughput from #178. That was a 16
MiB /
300-file smoke run - a pipeline proof, not a verdict - and a headline
MiB/s
from it would not survive scrutiny. The two numbers that did make it in
are
locally measured and honest at their scale: the exclusion-preview
re-classification (536 ms cached vs 868 ms fresh walk on a 63k-entry
tree,
from #177) and the scan thread clamp. The bench suite appears as a row
on its
own merits, with the caveats left in `bench/README.md`.
## Competitor claims and where each was verified
Versions checked: rclone 1.74.4, restic 0.19.1, Duplicati 2.3.0.4,
Backblaze Personal Backup 10.0.2, Drive for desktop 128.0.
**CPU / I/O priority - nobody else has it.** rclone has no priority code
and
closed both requests pointing at `ionice`
(rclone/rclone#864). restic's FAQ answers "How
to
prioritize restic's IO and CPU time" entirely with `ionice`/`nice`
recipes
(https://restic.readthedocs.io/en/stable/faq.html). Duplicati still
accepts
`--thread-priority` but the shipping string is "has no effect, use the
operating system controls to set the process priority"
(`Duplicati/Library/Main/Options.cs` at the `v2.3.0.4_stable_2026-07-09`
tag)
- note its published docs page still lists the old text with no
deprecation
notice, so I cited the source, not the doc. Drive for desktop's entire
preference surface is bandwidth rate limits plus pause
(https://support.google.com/drive/answer/13470231), with no priority key
in
the admin policy list. Backblaze offers an automatic/manual bandwidth
throttle
and an upload-thread count
(https://www.backblaze.com/computer-backup/docs/configure-performance-settings-windows).
Bandwidth limiting is not I/O priority and the note says so.
**Parallel scan.** rclone walks at `--checkers` (8) and transfers at
`--transfers` (4) (https://rclone.org/docs/). Duplicati's
`FileEnumerationProcess` is a single serial task with the concurrency
downstream of it, so its walk is a genuine `:x:` (source at the stable
tag).
restic reads at `--read-concurrency`, default 2
(https://restic.readthedocs.io/en/stable/manual_rest.html). Google and
Backblaze document nothing about scan concurrency - hence the new
"not documented" marker rather than a guessed `:x:`.
**Destination-side deletion.** rclone gets a `:white_check_mark:` here,
not an
`:x:`: it keeps no state, so `copy`/`sync` re-list the destination every
run
and re-transfer anything missing
(https://rclone.org/commands/rclone_sync/).
The note says so plainly, including that this is why it is slower on the
incremental case. Drive for desktop is the one that is worse than absent
-
"any files you put in the trash are put in the trash everywhere"
(https://support.google.com/drive/answer/2375102). Duplicati and restic
detect
damage but recovery is operator-driven (Duplicati's `RepairHandler.cs`
refuses
missing dblock files without `--rebuild-missing-dblock-files`; restic's
troubleshooting doc says re-run `backup` to heal,
https://restic.readthedocs.io/en/stable/077_troubleshooting.html).
Backblaze
documents nothing either way, so it is marked not-documented rather than
`:x:`.
**Live rule preview - and one claim I had to walk back.** Duplicati
shipped
server-evaluated inclusion state in its tree UI in 2.3.0.4 on 2026-07-09
(duplicati/duplicati#6955, closing a
six-year-old
request, duplicati/duplicati#4194). Six weeks
ago
"nobody else has this" would have been true; it is not any more, so
Duplicati
gets a `:white_check_mark:` and the note records the remaining
difference
(it marks the nodes you expand, with no whole-source counts). rclone and
restic are `--dry-run` only.
**Logs and diagnostics.** Driven is not unique here and the row shows
that:
Drive for desktop and Backblaze both get `:white_check_mark:`
(https://knowledge.workspace.google.com/admin/drive/capture-google-drive-for-desktop-logs-for-support,
https://www.backblaze.com/computer-backup/docs/send-logs-to-backblaze-windows).
Duplicati is partial - its "Create bug report" bundle is real, but file
logging is opt-in, warnings-only by default, and unrotated. restic has
no
log-file option at all.
**Benchmark suite.** Negative evidence for all five (repo tree listings
plus
code search); the note names Duplicati's unreleased AutoTune harness and
Backblaze's B2-not-client benchmark rather than pretending there is
nothing
adjacent.
No existing rows were removed and none had become false. The trailing
line
now names the exact versions checked instead of just the month.
## Checks
- 40 notes, sequential, every table marker resolves to one and every
note is
cited (validated by script).
- All 27 table rows have 7 cells.
- Zero em/en dashes or other dash-like non-ASCII; only the superscript
digits
the file already used. `git ls-files --eol` reports `w/lf`.
- README is not covered by prettier or any markdown linter in CI
(prettier
runs against `ui/src` only), so there is no formatting gate to run.
- `deploy-landing.yml`'s `TAGLINE_MARKER` is the README's first line,
which is
untouched.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
https://claude.ai/code/session_01JLB3E2Jm7knNJd37fVpH8X
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>1 parent 85f6d6a commit e669d57
1 file changed
Lines changed: 65 additions & 20 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
21 | 21 | | |
22 | 22 | | |
23 | 23 | | |
24 | | - | |
25 | | - | |
26 | | - | |
27 | | - | |
| 24 | + | |
| 25 | + | |
| 26 | + | |
| 27 | + | |
| 28 | + | |
28 | 29 | | |
29 | | - | |
| 30 | + | |
30 | 31 | | |
31 | 32 | | |
32 | 33 | | |
| |||
43 | 44 | | |
44 | 45 | | |
45 | 46 | | |
46 | | - | |
47 | | - | |
48 | | - | |
49 | | - | |
| 47 | + | |
| 48 | + | |
| 49 | + | |
| 50 | + | |
| 51 | + | |
| 52 | + | |
| 53 | + | |
| 54 | + | |
| 55 | + | |
50 | 56 | | |
51 | | - | |
| 57 | + | |
| 58 | + | |
52 | 59 | | |
53 | 60 | | |
54 | 61 | | |
| |||
65 | 72 | | |
66 | 73 | | |
67 | 74 | | |
68 | | - | |
69 | | - | |
70 | | - | |
71 | | - | |
72 | | - | |
73 | | - | |
74 | | - | |
75 | | - | |
| 75 | + | |
| 76 | + | |
| 77 | + | |
| 78 | + | |
| 79 | + | |
| 80 | + | |
| 81 | + | |
| 82 | + | |
| 83 | + | |
| 84 | + | |
| 85 | + | |
| 86 | + | |
| 87 | + | |
| 88 | + | |
| 89 | + | |
| 90 | + | |
| 91 | + | |
| 92 | + | |
| 93 | + | |
| 94 | + | |
| 95 | + | |
| 96 | + | |
| 97 | + | |
| 98 | + | |
| 99 | + | |
| 100 | + | |
| 101 | + | |
| 102 | + | |
| 103 | + | |
| 104 | + | |
| 105 | + | |
76 | 106 | | |
77 | 107 | | |
78 | 108 | | |
79 | 109 | | |
80 | 110 | | |
81 | 111 | | |
82 | 112 | | |
83 | | - | |
84 | | - | |
| 113 | + | |
| 114 | + | |
| 115 | + | |
| 116 | + | |
| 117 | + | |
85 | 118 | | |
| 119 | + | |
| 120 | + | |
86 | 121 | | |
87 | 122 | | |
| 123 | + | |
| 124 | + | |
| 125 | + | |
88 | 126 | | |
89 | 127 | | |
90 | 128 | | |
91 | 129 | | |
| 130 | + | |
| 131 | + | |
92 | 132 | | |
93 | 133 | | |
94 | 134 | | |
| |||
206 | 246 | | |
207 | 247 | | |
208 | 248 | | |
| 249 | + | |
209 | 250 | | |
210 | 251 | | |
| 252 | + | |
| 253 | + | |
| 254 | + | |
| 255 | + | |
211 | 256 | | |
212 | 257 | | |
213 | 258 | | |
| |||
0 commit comments