Date: Sat, 13 Dec 2008 18:17:25 +0300 From: Boris Samorodov <bsam@ipt.ru> To: Palle Girgensohn <girgen@pingpong.net> Cc: Tony Jago <tony@convoitec.com>, Alec Kloss <alec@setfilepointer.com>, "<freebsd-afs@freebsd.org>" <freebsd-afs@freebsd.org>, "Jason C. Wells" <jcw@highperformance.net>, Derrick Brashear <shadow@gmail.com> Subject: Re: OpenAFS port Message-ID: <46367482@bb.ipt.ru> In-Reply-To: <C189502E01D49971EB598087@girgBook.local> (Palle Girgensohn's message of "Sat\, 13 Dec 2008 15\:15\:19 %2B0100") References: <493ACAC4.5020806@linuxbox.com> <12501719@bb.ipt.ru> <493D898C.1030609@linuxbox.com> <22B6C509EF7C4AB0A2D8350C31BB8D5D@valentine> <57098597@bb.ipt.ru> <26695644@bb.ipt.ru> <DC87E29101195307B372C4F5@c-3157e155.1521-1-64736c12.cust.bredbandsbolaget.se> <20081213004251.GA88954@keira.kiwi-computer.com> <db6e3f110812121706i2b022e0bh3ff7413086c73dc1@mail.gmail.com> <A22DDF0293864B03AD8FE957D5EB5316@valentine> <42451957-717C-4CA3-97D9-E2ACABE55E34@pingpong.net> <60600083@bb.ipt.ru> <C189502E01D49971EB598087@girgBook.local>
next in thread | previous in thread | raw e-mail | index | archive | help
Palle Girgensohn <girgen@pingpong.net> writes: > --On lördag, lö 13 dec 2008 16.19.40 +0300 Boris Samorodov > <bsam@ipt.ru> wrote: >> Palle Girgensohn <girgen@pingpong.net> writes: >>> 13 dec 2008 kl. 03.27 skrev "Tony Jago" <tony@convoitec.com>: >>> >>>> I think that we probably don't need more then one port. Yes, I know >>>> I was the one what originally proposed the meta port but I have >>>> changed my mind :) The reason we had a server and a client port The more I work on the port the more I understand that it's not so trivial. I've changed my mind too. ;-) >>>> originally was that the server was the only bit working and the >>>> kernel model was set not to compile. The client was was arla client. >>>> Now that both the openafs server and client are supported by the >>>> openafs team I can see no reason why it shouldn't be all in one >>>> port. The port should have separate rc variable to allow the >>>> administrator to only start the client or the server if they choose >>>> to. openafs_client_enable="YES" and openafs_server_enable="YES" for >>>> example. This gets around all the conflicting file problems. The >>>> kernel module need only be loaded if the client is required. This >>>> would seem to be a much easier and cleaner solution. >>> >>> As long as nothing conflicts with arla, I also suggest an all in one >>> installation. Keeps it simple, which is always important. >> >> I'd vote for that myself if and only if we speak about a ports >> subsystem. But there are packages as well. And for those who prefer >> using packages I'd rather give an opportunity. >> >> Said that I propose following ports: >> . net/openafs (server+client) >> . net/openafs-server; >> . net/openafs-client. >> >> One of them will be a master port (I don't figure out which one, >> but that will be either openafs or openafs-server). All of them >> will conflict each other, i.e. only one of them can be installed >> at a machine. >> >> That way we may give all users their chance. >> >> Opinions? Thanks! > > Personally I thinks that is overly complicated. Installing a few > binaries that I don't use is not a big deal. Having three ports for a > server+client system is confusing, especially since it is not very > common practice. I'd vote for one single port, with rc switches to > activate the different parts. Anyway we can't create packages since kernel sources and objs are needed. Then really it's worth having only one port. The default net/openafs installs both server and client. While OPENAFS_SERVER_ONLY and OPENAFS_CLIENT_ONLY do what they should. WBR -- Boris Samorodov (bsam) Research Engineer, http://www.ipt.ru Telephone & Internet SP FreeBSD committer, http://www.FreeBSD.org The Power To Serve
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?46367482>