What would the benefits be having the proxy do gziping Vs your application?

What would the benefits be having the proxy do gziping Vs your application?

It would be simpler to enable and I guess the gzip would be faster in a Go app (i guess the proxy is made in Go) than doing that in javascript + reduce any overhead that gzip may have in my application.

Might as well throw in brotli and zstd compression for good measure

compression is pretty easy to get started, I can add gzip right now. But then when we start to introduce other compressions like brotli and zstd, there are latency trade offs which are not good for Railway user's since we don't cache anything at our edge so this latency cost is hit every time. We'd have to develop some heuristics to know given the content-type and accepted-encoding, which compression to use. We'd also want to allow user's to configure this themselves for their application in case our default settings are not good.

If we want to add decompression (client uploads something compressed) then it gets much trickier so we'll save that for another thread.

I have something that implements gzip level 5 and brotli level 3. These are conservative levels (closer to 1 is faster with less compression). Should be able to merge tomorrow. Brotli is prioritized over gzip.

This is also just client to proxy. Not e2e.