ICRC3, Archive functionality. Are timers necessary?

Developers
1

I am trying to implement the archive functionality for ICRC3

I was learning form the ICDev implementation: GitHub - PanIndustrial-Org/icrc3.mo: ICRC3 in motoko

In this implementation, they start to archive the ledger state by starting a timer with a period of 0 seconds:

f(Vec.size(state.ledger) > state.constants.archiveProperties.maxActiveRecords){
            state.cleaningTimer := ?Timer.setTimer<system>(#seconds(0), check_clean_up); '

Inside the timer function “check_clean_up” , they do everything necessary to create an archive canister: create an instance of archive canister, give cycles, and stop the timer once all ledger blocks are archived.

I cannot understand why they use a timer. Why can’t they just call a function?

2

add_record is a synchronous function and we did not want to convert it to async so that it could be used in other sync functions. The timer makes the clean_up run out of band so that the user sending a transaction doesn’t have to wait for potentially a new canister to be created and then records sent to it.

If we had called it directly we would have had to make the return type async.

3

What is the motivation of doing add_record synchronous? Do you want to make sure that the record is introduced inthe ledger before the client gets a response? if this is the case, why?

4

Synchronous is almost always better from a atomicity point of view.

We want to return to the user quickly and the archiving is an offline operation that no one needs to wait for.

With ICRC-4 you may hit the archive threshold much more often(sometime maybe even twice in the same transaction). Doing in a different execution context lets the archivers settle it self out during high volume situations.