Poul-Henning Kamp
Poul-Henning Kamp
How does this work with H2 ? I guess it causes head-of-line-blocking, since I see nothing tweaking the H2 windows to stop the client from pushing ahead ? Do we...
Thinking about this, and our approach to possible-DoS-situation logging: Maybe we should log details for the first incident, then start at timer of one minute ($param) and not report details...
bug-pow-wow sorta-consensus: Maybe we need SC_SESSION_WS_OVERFLOW, SC_CLIENT_WS_OVERFLOW etc...
Split SC_OVERLOAD into different ones which say what we ran out of
Originally we decided "to not leak information" and limited error responses to 503 and one or two 40x responses. I agree with @Dridi that we need to decide how compliant...
Bugwash: Not a 2020-march-15 subject, needs either VCL4.1 or V7.0, possibly both. @dridi to open VIP on subject of greater compliance for 2020-sep-15, everybody to help out.
Until now we have not done this, under the assumption that users wanted the VSL records to precisely report the entries in the acl.
I think "prune" is a better name than "merge" ? Given that ACL's no longer VSL log by default, it is not obvious to me if this optimization is worth...
I'm very paranoid about our ACL's being correct, so I did a couple of manual tests, and there seems to be an "off-by-8" kind of error: "10.0.0.0"/23; "10.0.2.0"/23; "10.1.0.0"/24; "10.1.1.0"/24;...
We do not want the parameters to be "4kb + overhead" because that would be suboptimal for VM reasons. We should probably set a "dynamic minimum" on these params to...