Search was still stalling for seconds a keystroke. A trace across five keystrokes said
the renderer main thread was blocked for 9.4 s in a single task with *zero* V8 samples
inside it - no JavaScript ran at all. What ran instead, on the worker pool during those
same 9.4 s: ImageDecodeTask 7.6 s, RasterTask 1.7 s, and only 229 ms of actual "Decode
Image". The main thread was waiting on cover bytes, not computing anything.
Two causes, and the first one was mine. The stale-while-revalidate handler added earlier
today made every cover do a Cache Storage read *plus* a network fetch *plus* a cache
write, all serialised through one worker thread. Measured with
Network.setBypassServiceWorker: worst keystroke 5580 ms through the worker against
461 ms without it. So covers no longer go through the worker at all. That costs nothing
here: this worker only registers on localhost or over HTTPS, which on this setup is the
kiosk on the device itself, where the backend is the same machine and a cache lookup is
strictly more work than asking for the file. The shell caching, which is what makes it
installable, stays.
The second is that decode cost goes with pixel count, and 640 px was headroom for a
tablet at devicePixelRatio 2 that nobody had asked for. A browse grid paints a card
132 px wide; the largest any screen asks for is 340. At 384 px, typing "conni" over 343
albums, keydown to painted, three runs:
before 3735, 270, 5580, 88, 57 ms
after 873, 75, 17, 24, 26 ms
`shrink_cover` never scales art up, so dropping the limit cannot be applied by
re-reading the cache - only by going back to the original art. _INDEX_VERSION 6 does
that: the index is discarded and every album is scanned again through store_cover.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Shrinking the covers changed nothing on the device, twice over, because two layers were
independently serving the old bytes and each had the same wrong premise written into its
comment: that a cover is "content-addressed by album id". The id addresses which album
the art belongs to. The bytes behind the URL change whenever the art is reprocessed.
The backend sent `public, max-age=604800`, so every browser that had loaded the page in
the previous week kept decoding the 3000px original out of its own disk cache without
asking. It now sends `no-cache`, which does not mean "do not store" but "revalidate
before reusing" - the file stays cached and the usual answer is a 304.
The service worker cached covers cache-first on a URL that never changes, which means a
client could serve a stale cover for ever; bumping the cache name fixed today's covers
and would have had to be done again for the next batch. It now does
stale-while-revalidate: the cached copy is returned immediately, so a grid of album art
still never waits on the network, and a fresh copy is fetched behind the page for next
time. Neither layer needs a version bump when art is reprocessed again.
The revalidation is off the paint path entirely, which is what makes `no-cache`
affordable here: the service worker answers first, the network call happens behind it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A DevTools trace recorded on musicdolphin put the whole question beyond argument:
ImageDecodeTask 29,939 ms x20 (~1.5 s each)
Decode LazyPixelRef 3,924 ms x20
Decode Image 3,530 ms x22
Paint 99.7 ms
Layout 77.3 ms
all app JavaScript <400 ms across 22 seconds
Image decoding is not the largest cost on that page, it is very nearly the only one.
The covers should already have been small - the backend downscales to 640 px now - but
the page was still decoding them at 1920, 1600, 1400 px, with deliveryType
"cache-storage" on every one. The service worker caches cover art cache-first, keyed on
URL, on the premise that it is "immutable per album id". The id did not change; the
bytes did. So every client that had ever loaded a cover kept serving the 3000 px
original from its own disk and never asked the backend for the new one.
Renaming the cache is the retirement mechanism the worker already has - `activate`
deletes every cache that is not one of the two current names - so COVERS becomes v2,
with a note saying that changing how covers are produced means bumping it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
