Er, sorry that didn’t sound right. Agree with you that not updating max-age
seems more like a bug if we are to go with the hypothesis.
> On Mar 26, 2021, at 3:02 PM, Sudheer Vinukonda
> wrote:
>
>
> Thanks Brian!
>
> Hmm, it seems to me that the traces you shared actually further confirm t
Thanks Brian!
Hmm, it seems to me that the traces you shared actually further confirm the
hypothesis - ie ATS serves a locally cached stale object because of negative
revalidation (and Origin connection error/5xx), yet we do *not* want the
clients receiving that copy (whether it's an actual br
Thank you Sudheer. This is really helpful. That at least gives some
justification for the current behavior: try to communicate the lifetime of
the object to something in the future so the upstream clients or proxies
don't think they need to revalidate their cached objects. It is
insufficient, unfor
Hi Brian,
Thank you for the detailed and excellent write up on the current behavior!
Just to play a bit of devil's advocate, I'm wondering whether the
`negative_revalidating_lifetime` is being interpreted incorrectly. I do admit
it's confusing given the name and the documentation (which can bot
