Jonas Wallden <<jonasw[at]roxen.com>> wrote:
> True, but to be honest I didn't introduce Standards.URI aside from a
> single place; /.../
Yes, sorry, I was unclear. What I meant was that your patch started
using Standards.URI to format the urls, whereas the old code used some
> Maybe an option in Standards.URI to coerce to a string form including
> port numbers?
That'd be the nice way to solve it, yes. It'd introduce a hard
dependency between pike and roxen, but that's a transient problem.
Don't have any obvious ideas on how to extend Standards.URI in a neat
> Reverting my changes will likely break IPv6 adresses.
Yes, they wouldn't really be reverted, just the url formatting bits.
Are there any ipv6 specials there, so that a simple
sprintf ("%s://%s:%d%s", protocol, host, port, path)
> Otherwise dropping the port number when constructing the parameter
> to find_configuration_for_url() could be good enough?
Regardless how it's done, that'd introduce more overhead in the hot
code path instead. Even though it still might be small compared to all
other work, I still think it's wrong when the problem can be solved