Nils Goroll

notes from bugwash: > (15:39:32) slink: I am back to thinking that just having headroom_xxx and auto-calculating the workspace_xxx would be the simplest and cleanest option > (15:39:59) slink: workspace_xxx...

FTR: This has been discussed during bugwash with no clear decision. Some notes: * A related question is if `keep` should be active at all for non 304-able objects. It...

This ticket has been discussed between @bsdphk and myself. Brief summary: We did not come up with a particularly nice solution, any ideas to clean the object list in retrospect...

> an extra lookup at the end ... is not what I meant. I meant to move a found stale_oc to the end of the oc list such that another...

note to /self: test on SunOS

> SunOS: `make check` in `contrib` passes as we just found out, only because it hits GNU awk first. It fails with `/usr/bin/awk`: ``` ... ** top === shell {...

3d331ab2961d1de1882f9a9b9aabe3ef54c4cec6 works with `/usr/bin/awk` on SunOS: ``` $ strings /usr/bin/awk | grep -E '^vers' version Oct 11, 1989 ```

@bsdphk yes to the first question: The scenario is to modify the headers of the cache-object with respect to the downstream behavior, but have a differing ttl/cacheable status. And yes,...

Discussed on IRC: I think this model should work: * before `vcl_backend_response {}` , for a 304, call `vcl_backend_refresh {}` with `beresp.*` (r/w) and `obj.*` (r/o) * `builtin.vcl` would just...

@Dridi how would you handle a misbehaving backend when you already have a 304 response (with no body)? I do not see how `return (replace)` vs. `return (reuse)` would work,...