Re: Feature to dynamically disable PSL dataset from filesystem

> Would it perhaps be better to work on making the load/refresh happen at a
different moment in time than exactly when parsing the cookie?
> Since the refresh timeout is 72 hours and few applications actually live
that
long, it seems to mostly be a problem with the first load. Is that perhaps
so?

I explored sources of userver a little bit,  under the hood there is pool
of easy handles, this pool has application lifetime and allows to reuse
easy handles, so prefix graph can be reloaded again after 72 hours.
I like your idea to move psl initialization at some other place, as
improvement I like the idea to have a public way to warmup psl cache with
user supplied expiration timeout (which can be infinite or big enough?)
which will allow to select place where to warmup psl and how long to keep
it (userver has dedicated task manager for IO bound operations, such fs
blocking operations are executed there). Of course this warmup is fully
optional to delegate psl initialization with own expiration time to
libcurl, because I suppose not so many users want to deep in that.

P.S. I thought about the way to mock psl interface. But it can noise API,
maybe it is not best way for such exotic feature.

-- 
Unsubscribe: https://lists.haxx.se/mailman/listinfo/curl-library
Etiquette:   https://curl.se/mail/etiquette.html