Reviewing entries after the recent standard mode clarifications on #1068 / #1069 ("for standard we must use the default compression middleware instead of a hand-rolled version", no custom levels, from the review of the express and fastify entries):
The elysia entry is marked mode: standard but negotiates compression by hand in /json/:count, calling Bun.gzipSync directly (server.ts).
Separately, its br branch has a bug: it sets content-encoding: br while compressing with Bun.deflateSync, so those are deflate bytes under a brotli label. It is latent today because validation sends Accept-Encoding: gzip, br and the code always picks the first token, so the branch never runs there.
Happy to send a PR moving the entry to a compression plugin with default settings, or it could be re-labeled tuned; the br mislabel deserves a fix either way. Asking mostly for consistency: the same rules were (rightly) applied to the entries I submitted, and the composite ranking compares these rows directly.
Reviewing entries after the recent
standardmode clarifications on #1068 / #1069 ("for standard we must use the default compression middleware instead of a hand-rolled version", no custom levels, from the review of the express and fastify entries):The elysia entry is marked
mode: standardbut negotiates compression by hand in/json/:count, callingBun.gzipSyncdirectly (server.ts).Separately, its
brbranch has a bug: it setscontent-encoding: brwhile compressing withBun.deflateSync, so those are deflate bytes under a brotli label. It is latent today because validation sendsAccept-Encoding: gzip, brand the code always picks the first token, so the branch never runs there.Happy to send a PR moving the entry to a compression plugin with default settings, or it could be re-labeled
tuned; thebrmislabel deserves a fix either way. Asking mostly for consistency: the same rules were (rightly) applied to the entries I submitted, and the composite ranking compares these rows directly.