On Jun 5, 2008, at 8:04 AM, Laurent Blume wrote: >> They will be left behind. The choice was to move >> forward. > > Blindly, without looking what others are doing or needing? I'm > really curious as to what kind of consideration was given to do the > transition from one model to another? I know that there has been > some work in the Linux world to do POSIX/NFSv4 mapping, was that > possibility rejected for Solaris?
Of course it wasn't done blindly. The NFS community is one that communicates regularly and civilly. Andreas Gruenbacher has worked on NFSv4 ACLs for Linux off/on for awhile. One presentation I found of his is here: http://www.suse.de/~agruen/agruen-nfs4acl.pdf Andreas and Bruce Fields have engaged the NFSv4 community in an effort to work out the issues of mapping of ACLs and there was clarifying language added to the NFSv4.1 Internet Draft as result. All of the issues related to mapping haven't been resolved because of an interpretation difference related to the Posix standards (I should point out that this reference is to the main Posix standard and not the Posix Draft that contains ACLs). As to the status of Andreas' work, I don't know. He will have to speak for himself. >> >> At some point a decision has to be made to move >> forward. >> In the world of NFS, this has usually meant that >> previous versions can be served and in this one case, >> over the last 20 years, an inconvenient >> incompatibility >> has been created. Of course, in those last 20 years, >> there hasn't been a standard ACL format so the issue >> is really moot. > > The fact that there was, regrettably, no standard doesn't mean > there was nothing. Like people were already using word processors > before ODF became a standard, and ODF didn't suddenly become a > worldwide success because it is a standard. > I have those non-standard ACLs working right now on Linux systems, > and they will stay for a while, whether I like it or not. Sure because the underlying filesystems will be present. > > I have the feeling that Sun didn't really try to convince others to > follow their steps, and just assumed the standard would > automatically attract them. I hope I'm wrong. You are. Sun, Netapp, EMC, AIX have chosen to provide NFSv4 ACLs natively. It is the end users that will have the most voice in convincing implementors to make a choice. Which brings us to your original problem and how OpenSolaris can help solve it. Let's review that again... Spencer
