Open irori opened 5 years ago
In the long run, I think it'll be useful to support encoded inner responses in bundles to optimize local disk storage. Since signed exchanges don't need to be random access, it's less important to support compressing the inner response if that makes the implementation trickier.
In the short run, I think it'd be fine to say that seeing Accept-Encoding
in the Variants
header results in a mismatch
from https://wicg.github.io/webpackage/loading.html#request-matching. I won't get around to this in the next couple weeks.
I think it'll be useful to support encoded inner responses in bundles to optimize local disk storage. Since signed exchanges don't need to be random access
👍
it don't make any since to (de)encode things like video and images that are already compressed... So it would be wasteful to compress the hole web bundle
When fetching a Signed Exchange, Chromium sets
Accept-Encoding
request header togzip, deflate, br
. However, actually Chromium accepts differnt set of content encodings for Signed Exchange's outer and inner responses.gzip
,deflate
, andbr
are supported for outer responses, and onlymi-sha256-03
is supported for inner responses.This is problematic when performing the Request Matching. For example if exchange has:
It will not match the
browserRequest
which hasAccept-Encoding: gzip, deflate, br
.What should we do?
Ideas:
Accept-Encoding
header tomi-sha256-03
.