How does `fetch`'s cache behave with query strings?
How does
fetch
EIt respects whatever cache setting you have for the zone the worker runs on
Dis there a way to configure that per-fetch?
EI don't believe so, no
Dhmm, could I do a cache rule maybe?
EBut while we're here, it seems fetch supports
cacheTtlByStatusthat's enterprise-only; I know I'm gonna forget about it when we downgrade off ent eventually 
EAh
EOkay
Dwhat on earth, Backblaze returns a 403 when it can't find a file
Dis that actually the s3 spec...?
JcacheTtlByStatus
Djust confirmed, Backblaze B2 returns
when it's a 403 (missing file)
Cache-Control: max-age=0, no-cache, no-storewhen it's a 403 (missing file)
Jhttps://developers.cloudflare.com/workers/runtime-apis/request/#incomingrequestcfproperties doesn't mention this, but https://developers.cloudflare.com/cache/how-to/configure-cache-status-code/ kinda does in the first sentence. But then goes on to talk about free limits too. interesting
EYou can do it with page rules. I had thought there was an equivalent in cache rules but I checked and can't find it so...
It feels wrong to write a page rule for a hostname thats not on the zone but I'm 99% sure it works
It feels wrong to write a page rule for a hostname thats not on the zone but I'm 99% sure it works
DI think it's under cache key?
EOh, duh, its been modernised
EThanks
UCache key is ent only isn’t it
Dlemme confirm on my non-enter CF account
EOn free plan you get some options, just limited:
E
Ddang it I got ninja'd
UInteresting
EYou need ent for the rest
Elike specific headers, specific query params, etc.
UAh this is what I was thinking of
Cyou get query string sorting and origin passthrough for free via cache rules
Cnice guy cf giving ent stuff lol
EYuh, the cache rules stuff is basically a pretty UI for those custom cache keys. If you want to go crazy and write a string yourself you can do that in a worker (though its a lot easier to get it wrong and mess up your cache)
D
Ddoes this look sane if I want to cache all S3 API requests?
EAnd True-Client-IP with transform rules
Elgtm
EOutbound fetch requests go through cache rules and page rules, it feels weird to write them for external hosts but its powerful
Dyou get status code ttl caching too lol
Chuh I didn't even realize that was an ent thing
Dthis confuses me, which one takes priority?
Dwill it use cache-control if present, and if not, the status code ttl?
UI think you need the last option
UAlthough I guess that would be the input right under that’s very confusing
EI'm pretty sure status code TTL is an override, using it instead of "use this TTL" is a tad redundant
DI don't want to override non-200s though
EI think they would have the same effect, this one just makes more sense in my mind
E(note that due to the cache-control spec being super weird you have to do no-store and should not pick no-cache)
Dyeah that threw me off earlier today too
D
Dah, this is probably because I only gave the B2/S3 API key permissions to read and write files, not list files.
ENice, though we dont cache any 1xx by default and I'm pretty sure the TTL override at the top only applies to 200 :P
Either way, good to be explicit like that and its more clear in future what this rule is doing and can be easily adjusted if needed!
Either way, good to be explicit like that and its more clear in future what this rule is doing and can be easily adjusted if needed!
Dalso did this to avoid caching versioned presigned URLs.
