On Mon, Oct 10, 2022 at 11:55 AM Yann Ylavic <ylavic....@gmail.com> wrote: > > On Fri, Oct 7, 2022 at 9:14 PM Ruediger Pluem <rpl...@apache.org> wrote: > > > > On 10/7/22 7:11 PM, Stefan Eissing via dev wrote: > > > > > > > > >> Am 07.10.2022 um 18:45 schrieb Yann Ylavic <ylavic....@gmail.com>: > > >> > > > > > > Thanks, Yann, for the detailed explanation on how this works and that the > > > default does the right thing. > > > > +1. I missed the default of not reusing the connection in these cases but > > having the possibility to override it like the user did. > > But shouldn't we default to not reusing the connection in case of a $ > > inside the port as well? > > We default to disabling connection reuse for any $ substitution > already, the compat issue is for an explicit enablereuse=on which used > to "work" because it was ignored.. > > How about a patch like the attached one now, which disables connection > reuse definitively (regardless of enablereuse) if there is a $ > substitution in the hostname or port part of the URL? > The patch uses the existing "is_address_reusable" flag (set to false > initially in this case), and since it's not configurable by the user > we are safe from a connection reuse point of vue in any case.
+1 in concept