Date: Mon, 21 Aug 95 12:25:56 MDT From: terry@cs.weber.edu (Terry Lambert) To: steve@simon.chi.il.us (Steven E. Piette) Cc: freebsd-hackers@freebsd.org Subject: Re: Making a FreeBSD NFS server Message-ID: <9508211825.AA26197@cs.weber.edu> In-Reply-To: <m0skPcC-0006IIC@simon.chi.il.us> from "Steven E. Piette" at Aug 21, 95 00:41:00 am
next in thread | previous in thread | raw e-mail | index | archive | help
> As I said earler, This mail wasn't suppost to be sent. My mistake. I'm > sorry. What I was really responding to was a interpretation on my part that > you were saying, that Sun's do async writes by default, and what you just > did say, that SVR4 (in the generic case) does as well. As others have > pointed out Sun does sync writes by default and belives some sort of > NVRAM is required to support safe async writes in NFS today. No problem. I made the response last night, before your other post, so we're kind of interleaved on I/O ourselves. 8-). > I had put my response aside to refer to both the SVID and the pre V3 NFS spec > before commenting and I sure I would have revised my initial response > which is more akin to talking to myself that actually directed at you. In all fairness, I have to make the claim that SVID III is worse than useless. AT&T/USL/Novell UNIX has failed SVID III compliance for a long time now (though not failed compliance testing) mostly because of the getitimer/setitimer/gettimeofday. It's the main reason that their select() implementation sucks out. I tried for a year to get this corrected, going so far as to totally rewrite the UnixWare timer code and write a working QIC-40/QIC-80 tape driver that used the modified code to try and get them to accept it into their source tree. No dice. > I've since checked SVID 3 and it says nothing about the expected behaviour > of NFS writes, so I say that in SVR4 its an implementation detail. I haven't > today checked the V2 NFS spec as to what it says about async writes, so I > won't comment further on conformance. I agree that it's an implementation detail. I have my NFSv3 here, but my NFSv2 is at home right now, so I really can't comment authoritatively on compliance with v2 either. My memory tells me that async writes are non-compliant. > The result was that I snapped at you. And for that I apologize. Like I said, no problem. 8-). Terry Lambert terry@cs.weber.edu --- Any opinions in this posting are my own and not those of my present or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9508211825.AA26197>