Use:

text

a

b

0-173

google

wikipedia

Do not use:

text

.a..b.*://*.a/*://:

The importer removes surrounding dots, so .a. becomes a, but plain a is cleaner.

Exact matching behavior

The extension lowercases everything, so A and a are identical.

For each search-result hostname, it splits the hostname at dots and checks complete labels:

Indexed label Result hostname Hidden

a a.com Yes

a www.a.com Yes

a shop.a.net Yes

a data.com No

b b.org Yes

b www.b.org Yes

0-173 shop.0-173.net Yes

0-173 shop.example0-173.net No

You do not need separate rules for www, http, https, or paths. The extension reads the hostname from the result URL and ignores its scheme and path.

Firefox suspending the background script while idle is normal MV3 behavior.

---

pasted-text.txt

TXT

MV3 is the better choice for filtering, while MV2 could make long imports easier. The manifest version itself will not materially change the speed of IndexedDB label lookups or DOM-result hiding.

Concern MV3 MV2

Idle CPU and memory Better: background can stop while idle Worse: persistent background page may remain loaded

SERP filtering speed Essentially the same Essentially the same

Long imports Requires careful resumable design Easier with a persistent background page

Future compatibility Better Legacy and increasingly unsupported

Recommended here Yes No

Your architecture is appropriate for MV3 because:

The options page streams source files sequentially.

The background performs short IndexedDB operations.

Completed sources can be resumed after interruption.

Content scripts process only supported search pages.

The large label collection is never loaded into every SERP tab.

A persistent MV2 background page might avoid suspension, but it would not solve the major resource costs: storing 17.6 million records, importing them, and performing exact disk lookups. It could also keep more memory resident unnecessarily.

Keep MV3. The highest-impact optimizations are fixing the Pause button, reducing metadata writes, batching lookups, adding cleanup for abandoned generation databases, and handling public-suffix labels safely. Changing to MV2 would provide little filtering benefit and create a legacy-extension dependency.

Deep research is initiating

Sources

No sources found

2026.07.18 22:01: "Start new full import builds a replacement index while the current index remains active.

Resume import becomes enabled only after an interrupted or paused build exists.

Pause import becomes enabled only while an import is actively running.

Delete active index removes the active index and disables effective filtering until another import completes."

You do not need to start another full import merely to test Pause. Your completed index is usable now. The next time you intentionally rebuild it:

Click Start new full import.

Once downloading begins, Pause import should be enabled.

Clicking it should stop the current download while preserving completed source files.

After pausing, Resume import should become enabled and the other button should change to Discard unfinished build.

Your completed active index remains available during a replacement import and after pausing one. Do not click Delete active index unless you deliberately want filtering to stop.