Summary of changes:

[quinnj]

Summary of changes:

• When we perform the HEAD request, and the output type is a buffer, we know what the final size should be, so allocate the final buffer up front
• We then pass "views" of that buffer to HTTP according to the specific Content-Range that a given task is requesting, this further avoids unnecessary allocations by having each HTTP request write directly to the slice of the final buffer it is responsible for
• Also added a debug message that will log the overal bit-rate achieved while downloading; useful for benchmarking
• Support passing in a pre-allocated byte buffer that will be used directly when writing the response body
• Allow passing keyword arg zlibng=true to use CodecZlibNG instead of CodecZlib for compression/decompression

[Drvi]

Also, if we rely on any new features introduced in HTTP, might be worth bumping the compat bound (e.g. for buffer reuse)

[quinnj]

Also, if we rely on any new features introduced in HTTP, might be worth bumping the compat bound (e.g. for buffer reuse)

Very good call.