On 11/5/21 11:00 AM, Jason Spashett wrote: > On the subject of unique ids, the code %random isn't necessarily of my > choosing. But it has 54^8 combinations, The number of unique values and the period of the random number generator are largely irrelevant here. What matters here is how soon the generator will repeat a _single_ value. Many random number generators repeat too soon, but most random number generators descriptions do not mention that fact because it is irrelevant for most things random numbers are (designed to be) used for. I do not know how soon your random number generator will repeat a single value, and I would worry about that if global uniqueness is the goal. > I think you are suggesting that a "proper" UUID may be better What is "better" depends on your desired uniqueness scope. If one does not care about cross-instance uniqueness and kid restarts, then an ever-increasing 64-bit integer counter is the best for them. This is similar to what %master_xaction provides. > Given that %master_axtion is a counter which resets when you restart > squid, and is not unique among squids (or restarts), is there not a > case to be made for making one available? Yes, there is absolutely such a use case! It can already be addressed via external ACLs, but a quality pull request adding native UUID support to globally-uniquely identify Squid processes (%process_uuid) should be welcomed IMO. The combination of (slowly computed once per process) %process_uuid and (fast) %master_xaction ID would give many admins what they want. Cheers, Alex. _______________________________________________ squid-users mailing list squid-users@xxxxxxxxxxxxxxxxxxxxx http://lists.squid-cache.org/listinfo/squid-users
