Tim Condon

Vapor takes a `TLSConfiguration` and passes it directly to NIOSSL (https://docs.vapor.codes/advanced/server/#tls) so it looks like we should support it out of the box as soon as it's released

Sorry for the delay. So I think this is the default behaviour for memory management on Linux in that the runtime won't release the memory unless the OS forces it...

Here's an excellent write up from @weissi about how the allocator works and how to see if it's actually leaking or not https://forums.swift.org/t/swift-rss-memory-consumption-is-much-greater-than-heap-memory/58837/10

The reason why it's not turned on by default in the Vapor template is it's not the default so is a bit of an unknown for us - I have...

After some investigation, it's caused by `GuardMiddleware` instantiated an `AbortError` which ends up in `StackTrace.capture`. On macOS this hits ```swift return Thread.callStackSymbols.dropFirst(1).map { line in let parts = line.split( separator:...

This is obviously something that's going to have to wait for the next major version of Vapor. Will leave it open

Yeah if you can make it work that would be great

(This should be added to the docs)

I'd be interested to see what other frameworks do in terms of picking the source of truth for the file type. Should the header or extension take precedence if both...

I wonder if there's something we can do around encoding them correctly/differently before passing them to `setenv` - I'll investigate