H
FPort that the server is listening on. 443 should be correct
Hops
H(Unauthorized) error: Service Error Message.; XML Error Message: <?xml version="1.0" encoding="UTF-8"?><Error><Code>Unauthorized</Code><Message>SigV2 authorization is not supported. Please use SigV4
EThe error comes from the CA, I will have to do some more extensive testing later within working hours unless someone else beats me to it
EYeah SigV2 isn't supported, what's trying to use it and does it support v4?
Dcan we constrain the allowed file types uploaded to r2 via presigned urls?
KIn S3 there is a conditions field which can limit what kind of files (size, content type, etc.) can be uploaded. I don't know if R2 supports this, the documentation says nothing about it.
DI'll give it a test, I know contentLength works, but not sure about the rest
KYou can presign the Content-Type header, but clients can just lie about what they’re uploading. If you need real content validation you need to do it on your server either before or after the upload
Dhow would you do it before?
Dafter would work but the lack of events makes this harder to implement
KIt depends on the uploaded file size. You could create an upload endpoint on your server and after checking the file upload it from your server to R2. For small files like images thats a way, but will get impractical for large files.
But all in all you should make a risk assessment what a malicious user achieve when manipulating the upload.
But all in all you should make a risk assessment what a malicious user achieve when manipulating the upload.
KOh and one other solution I've implemented once in a Nextjs app with large files:
Create an API endpoint where users upload file chunks to and keep track of the progress. Once all chunks where received (and forwarded to R2) I've known that the file upload was complete and I could send out a message to process the file.
Create an API endpoint where users upload file chunks to and keep track of the progress. Once all chunks where received (and forwarded to R2) I've known that the file upload was complete and I could send out a message to process the file.
Dah thats pretty cool
Dwe might opt for that approach depending on the complexity of handling file chunks
Dhonestly if r2 had events, it can be much more powerful and a real competitor to s3
KThats rather simple. You split up the file in chunks on the client and send it one by one
KI go long ways to not use S3 in any projects. 
Dright now, the main value proposition is pricing 
DI'm on GCP so it would be GCS for me haha
Was long as you pay for your usage nope.
Windeed, i said usage not egress :p
WWthe prices listed on that page yes
EGood question
Wtell 'em all to move to R2 
Ř
PHi sorry to interrupt but any fix for this?
KNope - you would need to read the file and then reupload it with the new content
Ucreate a new object and then list and read all 
EEven AWS S3 doesnt support it, there's no API mechanisn for append afaik
RHello. I've noticed an unexpected behaviour when creating R2 API Keys. Seems that API keys are bounded to each profile (user account) but not to the organization account?
This is an issue in case a developer that has created API keys leaves the company then we will lose immediate access to R2. Does someone know how to create an account wide API key?
This is an issue in case a developer that has created API keys leaves the company then we will lose immediate access to R2. Does someone know how to create an account wide API key?
KThat's how all API tokens work in Cloudflare, they're all associated to the user that created them.
KYou get 10 GB-month of free allowance every month. If you store 10 GB for 30 days, that is your free allowance.
KIf you stored 20 GB for 30 days, you'd only be billed for 10 GB-month.
CIs there a way to check if an object in an r2 bucket exist before making the read request
CI’m assuming no
KYou can do
HEAD, that's about it - but I doubt it's any faster than a GET returning a 404.KBoth will go to the same place and return the same stuff.
CYeah that’s what I thought 
trying to come up with a solution to tell users what objects are available without creating a separate database
trying to come up with a solution to tell users what objects are available without creating a separate database
CYeah the issue is that it incurs a read unit each time a nonexistent object is requested where as if the object does exist it is most likely already cached in cdn and won’t incur read
K404's are cached on the CDN for custom domains