Poul-Henning Kamp
Poul-Henning Kamp
after march15 release
Obviously this is a good idea. I have thought about the general shape of the filter-API, and checked some old notes, and I think I come down on the side...
I deliberately said "API" and not "ABI": Recompilation is cheap, stuffing source-code with #ifdef is not. Neither I nor Occam, share your passion for "multiplying entities" and I really do...
I think this is a valuable prototype, but it does have a lot of "bolted on" aspects which I would prefer to avoid. As mentioned during bugwash, originally the Mgt/Child...
I agree that this needs to always be under VCL control. On the other hand, I'm not keen on this becoming more complicated than it already is :-/
My view is that the IMS header indicates to VCL that conditional fetch is available, but nothing prevents VCL from replacing it with If-None-Match or another "better" conditional, for whatever...
I get why specifying the precise ban as arguments makes sense in general, but I think we should sugar the CLI to make it easy to cancel the most recently...
Forever is a long time, should that really be the behaviour ?
Bugwash consensus: 1msec is what we want for 0 timeouts. DocFix wanted.
This is not optimal. The problem is you dont want varnishadm to give up in interactive mode. Bugwash consensus: Give varnishadm ``-e`` argument (like shells) which makes it stop in...