General availability (i.e. not beta)

Greg pls let me in

Sorry, I'll show myself out

would have to be in the beta to show yourself out of it

thx

In theory yes? 🤷‍♂️ . I don't think we've gotten to testing that out yet. Presigned URLs are WIP at the moment which would be a prerequisite for something like that being useful I think, right?

I think it will be very useful

is there an R2 equivalent of S3's public-read acl?

Not for Open Beta initially.

true yes, cname is only useful for signed urls or public read buckets

fwiw, this is the S3 client library i'll be looking to use against R2's S3 endpoint - https://docs.aws.amazon.com/sdk-for-cpp/v1/developer-guide/examples-s3-crt.html (Uses multiple range requests per file to increase throughput)

Yeah, but they still only allow 1 range per get so you're net latency will still be the cost of 1 request. A true multi range read would actually be faster because you only pay the performance cost of the metadata lookup once and multiple ranges within storage nodes is trivial (& more efficient to be done in parallel within R2 than issuing multiple requests from the client). I've thought about it some more and I don't think that we can offer this for free. The good news is I think multiple ranges can be supported sooner rather than later provided we can make the billing easy to grok. The bad news is I'm not sure when, if ever, N ranges could be billed more cheaply than N GETs.

yeah, multi-range thing is interesting, but we've really only used range requests for the purposes of increasing throughput on large file fetches

in which case multiple GETs is really what you want anyway

@User - is there a way to make N ranges and set a limit on the total bytes + total ranges?

I ask because I fetch a lot of small pieces from larges files. I'm lucky that it's all small because I don't use unbound, so having to merge the pieces before sending them out as a single request myself had me scared with workers for a stretch.

I basically bunched all SDF glyphs into a single file, and have metadata so the client knows where to ask for individual pieces. The OG idea was to batch fetch in a singgle request to S3, but it turns out that literally no one supports multi-range requests.

Is the

r2_public_beta_bindings

compat flag live yet? 🙂

👀

https://canary.discord.com/channels/595317990191398933/940663374377783388/959567872449974352 🙂

Oh 😄

Missed that somehow.

If I'm understanding what you want this to mean is that you would provide N ranges and then want us to limit the total result to Y bytes? Not sure why you'd want this vs specifying the actual byte ranges you want. Can you help me understand? > I ask because I fetch a lot of small pieces from larges files. I'm lucky that it's all small because I don't use unbound, so having to merge the pieces before sending them out as a single request myself had me scared with workers for a stretch. I wouldn't recommend buffering and stitching in memory. Just create a TransformStream to do the stitching or a ReadableStream with a custom constructor. I'm not sure if this is with respect to the multi-range discussion that was happening earlier. If so, the resulting body returned would be stitched already.

I believe the edge is updated so if wrangler publish accepts the flag then you should be fine. It's possible we haven't updated the validator - it can lag a day or two. I've been out of town on business this week and will be back to regular hours next Tuesday.

Not sure if it has been asked already, but will we be allowed to store and serve video assets with R2? Cloudflare's Terms of Services limits the storage and serving of video assets to just the Stream product currently.

The supplemental terms changed recently - maybe for that purpose? https://www.cloudflare.com/supplemental-terms/#cloudflare-developer-platform

There used to be a specific part for Workers & whatnot that excluded videos specifically but that's gone.