You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently ACLs on public topics are prefixed on resource domainId + "._public". This isn't ideal, as it grants others access to topics that may not be in the spec.
E.g. given the spec contains topic <domain.id>._public.something and the cluster already contains, or someone adds later, topic <domain.id>._public.different, then anyone will be able to read that topic, even though its not managed by SpecMesh.
This is kind of an edge case, but we have all the data to be able to create exact literal ACLs, so I think we should.
Kind of related to #58. Maybe doing #58 is sufficient?
The text was updated successfully, but these errors were encountered:
my only concern with this idea is the cost of ACLs on broker performance. Once the topic growth has reached a certain point, then its difficult to change back to the current approach (i.e. remove granular ACLs and back to PREFIX only mode) - without operational overheads.
Currently ACLs on public topics are prefixed on resource
domainId + "._public"
. This isn't ideal, as it grants others access to topics that may not be in the spec.E.g. given the spec contains topic
<domain.id>._public.something
and the cluster already contains, or someone adds later, topic<domain.id>._public.different
, then anyone will be able to read that topic, even though its not managed by SpecMesh.This is kind of an edge case, but we have all the data to be able to create exact
literal
ACLs, so I think we should.Kind of related to #58. Maybe doing #58 is sufficient?
The text was updated successfully, but these errors were encountered: