From: "Vladislav Grishenko" References: <20200826165128.23117-1-themiron@...2010...> <027b01d6869f$2ce27910$86a76b30$@yandex-team.ru> <60d933c1-5238-1d0f-91e5-b94570180531@...1227...> In-Reply-To: <60d933c1-5238-1d0f-91e5-b94570180531@...1227...> Date: Wed, 9 Sep 2020 19:12:56 +0500 Message-ID: <02ba01d686b3$51a11ef0$f4e35cd0$@yandex-team.ru> MIME-Version: 1.0 Content-Language: ru Subject: Re: [Openvpn-devel] [PATCH v2] Add DNS SRV host discovery support List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable To: 'Arne Schwabe' , openvpn-devel@lists.sourceforge.net > From: Arne Schwabe > Sent: Wednesday, September 9, 2020 5:15 PM > > I see your point, thank you. > > Fallback is done just to conform RFC 2782 suggestion, personally I feel= it's > hardly be used with controllable client profiles when it's known that dom= ain.tld > does dns srv. >=20 > Yes. And see the point in that RFC. I would rather document that for out = fallback > the user has to manually specify that because I don't see a good "works f= or all" > fallback. Sure, I saw. RFC point of fallback was pretty valid for combined --remote f= or seamless migration dns->srv dns. With new --remote-srv this will be not necessary anymore, no migration here= . Moreover, facing the coexisting with usual --remote, fallback becomes pre= tty useless additional complexity. >=20 > > Additional --remote-srv indeed has benefits: > > * opens a non-conflicting syntax way to have multiple udp/tcp dns srv > requests, as you wrote in earlier mails. This will allow to sort them all= by > prio/weight. > > * cleaner syntax for specifying custom dns srv service name, with > > default "openvpn" and protocol (if need to limit discovery to only udp = or tcp) > So, in my mind syntax looks like: > > --remote-srv [service] [protocol], where both args are optional, by > > default service=3D=3Dopenvpn, protocol=3D=3Dany >=20 > Sounds good to me! Great, doing this way. >=20 > > And, important thing, multiple --remote-srv still needs to be supported= for: > > * custom order of protocol > > * different service names > > * allow to reuse --remote-random with --random-srv as well to achieve > > dns-srv-based lb > > > > Is it good enough in this form? >=20 > I don't really see the need for that but it doesn't break the normal case= of just > one remote-srv, so fine with me. >=20 > Arne >=20 -- Best Regards, Vladislav Grishenko