On Sat, 2022-11-05 at 09:47 +0100, Michael Osipov wrote: > Am 2022-11-04 um 12:59 schrieb Oleg Kalnichevski: > > On Fri, 2022-11-04 at 04:16 +0530, sreenivas somavarapu wrote: > > > >
... > > > > Sreenivas, > > > > The classic I/O (InputStream / OutputStream API based) is perfectly > > adequate and works quite well for request / response oriented > > protocols > > such as HTTP/1.1. > > > > The HTTP/2 protocol is completely different. It is frame based and > > can > > multiplex multiple message exchange streams over the same physical > > connection. The message stream multiplexing just does not work well > > with the classic I/O. Classic HttpClient 5.x implementation > > presently > > does not support HTTP/2 and most likely never will. If one needs > > HTTP/2 > > one has to switch to Async HttpClient 5.x. > > This also means that https://issues.apache.org/jira/browse/WAGON-606 > would never happen unless to be rewritten to the async client. > Yes, it does. I seriously doubt HTTP/2 support would bring any advantage to Maven Wagon. Two concurrent HTTP/1.1 connections will very likely do better as a transport for artifact upload / download than two concurrent HTTP/2 connections or two concurrent streams over a single HTTP/2 connection. Having said all that, how about this? I promise to build a HTTP/2 implementation of Maven Wagon and contribute it to the Maven project, if you can find some time to assess the state of our Kerberos code and help us decide if there is any hope of fixing it or we should just go ahead and deprecate it along with NTLM. Oleg --------------------------------------------------------------------- To unsubscribe, e-mail: httpclient-users-unsubscr...@hc.apache.org For additional commands, e-mail: httpclient-users-h...@hc.apache.org
