On Tue, 4 Jun 2024, Timothe Litt via curl-library wrote:
The application has no control over how many DNS transactions it takes to
resolve, e.g. a redirect/cname chain. So picking a limit would require
guesswork.
Correct. As a user you also cannot really know how many connections those DNS
transactions will need or create.
Still: the application has decided that it wants libcurl to limit the number
of concurrent connections and presumably such an applications do that for a
reason. So I don't think it is totally unreasonable for libcurl to then
actually try to adhere to that limit, even when DoH is used.
An alternate approach would be to not count the DOH connections at all - or
keep a separate cache and/or quota for them. You probably don't want a
long-lived libcurl user to keep DOH connections forever, or to cache very
large numbers. So a separate limit seems like the right approach.
Sure, but I am not a fan of inventing new complicated schemes if the much
simpler approach works good enough...
--
/ daniel.haxx.se
| Commercial curl support up to 24x7 is available!
| Private help, bug fixes, support, ports, new features
| https://curl.se/support.html
--
Unsubscribe: https://lists.haxx.se/mailman/listinfo/curl-library
Etiquette: https://curl.se/mail/etiquette.html
