Nils Goroll

@Dridi I believe my previous PR addressed all your concerns, except for the fact, that it used one additional lock on the same level as the `vcl_mtx`. I think that...

Preamble: I assume we agree that private headers can be implemented today to some extend (for a set of headers known a priori) by looking at `obj.hits` and changing the...

> Yes, you can actively cache a response that wouldn't fly by the built-in `vcl_backend_response` but without VMOD support it's hard to work with arbitrary headers. I have not come...

> > Can you please explain for which reason you want to put something into the object which has a lifetime exactly equivalent to that of the triggering request? >...

@hermunn with all due respect to your arachnal superpowers, we should design our software around the standard and leave special casing to special implementations, like a custom header sorting vmod.

@Dridi In that case `Accept-Language` should be merged. And yes, I think we all understand it's only a PoC, but in order to make progress, we should make decisions which...

In my mind, not asking for a conditional response should, under all circumstances, be more efficient for a backend than making a useless conditional request. In other words, if any...

> It might be better if you could explain why serving a 200 is faster than a 304? I am not saying that. I am saying that for an _empty_...

> I disagree, it might be cheaper to reply to a last-modified or etag request if they are for example indexed somehow on the backend side. Not having to run...

I regret having engaged with this ticket in the first place. I asked for a reference why an S3 backend should be faster for 304s than 200 with no body...