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 only one conceptual timer can be configured by filter instance by calling proxy_set_tick_period_milliseconds(). It implies that if multiple timers are required (i.e. multiple distinct tick_period's), the user has the responsibility to define an algorithm to decide the ideal tick period to cover all distinct tick_period's and writing the logic to demultiplex proxy_on_tick() on distinct listeners. Also the tick period lifecycle (i.e. when to erase or override the current tick_period) must be explicitly managed taking care of the multiple listeners. It seems a lot of boilerplate for a filter and looks more like an OS/platform responsibility. WASI clocks covers that use case, but it is probably far from what proxy-wasm can provide today, since it requires complex polling types that do not fit with the proxy-wasm event-loop style.
A more conservative approach may be to introduce a third parameter return_timer_id to proxy_set_tick_period_milliseconds() to return an identifier for the timer:
proxy_set_tick_period_milliseconds
params:
i32 (uint32_t) tick_period
i32 (uint32_t *) return_timer_id
returns:
i32 ([proxy_status_t]) status
Additionally proxy_on_tick() could receive the timer_id as second parameter:
proxy_on_tick
params:
i32 (uint32_t) plugin_context_id
i32 (uint32_t ) timer_id
returns:
none
With this API the user registers a new tick_period with proxy_set_tick_period_milliseconds(), receives a timer_id, and waits for the next proxy_on_tick() related to that timer_id.
The text was updated successfully, but these errors were encountered:
Currently only one conceptual timer can be configured by filter instance by calling
proxy_set_tick_period_milliseconds()
. It implies that if multiple timers are required (i.e. multiple distincttick_period's
), the user has the responsibility to define an algorithm to decide the ideal tick period to cover all distincttick_period's
and writing the logic to demultiplexproxy_on_tick()
on distinct listeners. Also the tick period lifecycle (i.e. when to erase or override the currenttick_period
) must be explicitly managed taking care of the multiple listeners. It seems a lot of boilerplate for a filter and looks more like an OS/platform responsibility.WASI clocks covers that use case, but it is probably far from what proxy-wasm can provide today, since it requires complex polling types that do not fit with the proxy-wasm event-loop style.
A more conservative approach may be to introduce a third parameter
return_timer_id
toproxy_set_tick_period_milliseconds()
to return an identifier for the timer:proxy_set_tick_period_milliseconds
i32 (uint32_t) tick_period
i32 (uint32_t *) return_timer_id
i32 (
[proxy_status_t
]) status
Additionally
proxy_on_tick()
could receive thetimer_id
as second parameter:proxy_on_tick
i32 (uint32_t) plugin_context_id
i32 (uint32_t ) timer_id
With this API the user registers a new
tick_period
withproxy_set_tick_period_milliseconds()
, receives atimer_id
, and waits for the nextproxy_on_tick()
related to thattimer_id
.The text was updated successfully, but these errors were encountered: