Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!news.bhp.com.au!mel.dit.csiro.au!munnari.OZ.AU!news.ecn.uoknor.edu!paladin.american.edu!europa.chnt.gtegsc.com!news.msfc.nasa.gov!elroy.jpl.nasa.gov!lll-winken.llnl.gov!venus.sun.com!news2me.EBay.Sun.COM!engnews2.Eng.Sun.COM!sunlite!pgt From: pgt@sunlite.Eng.Sun.COM (Panos Tsirigotis) Newsgroups: comp.unix.bsd.netbsd.misc,comp.unix.bsd.bsdi.misc,comp.unix.solaris,comp.unix.aix Subject: Re: ISP hardware/software choices (performance comparison) Date: 22 Jan 1996 19:31:28 GMT Organization: Sun Lines: 68 Distribution: inet Message-ID: <4e0omg$r7t@engnews2.Eng.Sun.COM> References: <4depms$bi5@park.uvsc.edu> <4dm8d8$30r@engnews2.Eng.Sun.COM> <4dp29j$8v8@park.uvsc.edu> NNTP-Posting-Host: sunlite.eng.sun.com Xref: euryale.cc.adfa.oz.au comp.unix.bsd.netbsd.misc:1987 comp.unix.bsd.bsdi.misc:2136 comp.unix.solaris:57372 comp.unix.aix:68802 In article <4dp29j$8v8@park.uvsc.edu> Terry Lambert <terry@lambert.org> writes: >pgt@sunlite.Eng.Sun.COM (Panos Tsirigotis) wrote: >] >Client caching prior to NFSv3 violates the protocol specification. >] >] Please post the relevant section of RFC-1094 (NFS spec) that supports >] the above assertion. > >|1. INTRODUCTION >| >| The Sun Network Filesystem (NFS) protocol provides transparent remote >| access to shared files across networks. >| > >"transparent". > >|1.3. Stateless Servers >| >| The NFS protocol was intended to be as stateless as possible. > >This is nice and tricky; the RFC is mostly a server spec, not a >client spec. > > >|2.2.8. Write to Cache >| >| void >| NFSPROC_WRITECACHE(void) = 7; >| >| To be used in the next protocol revision. > >Ie: not in this one. > I see that the misunderstanding stems from the fact that you interpret "transparent" the way you do, and assume that your interpretation is the only valid one. I would argue that NFS provides transparent remote file access, since a program accessing a file over NFS does not need to be concerned about the file's actual location, and it can access it using the same Unix API, as for local files. As a matter of fact, the designers of NFS give their definition of transparent access in "Design and Implementation of the Sun Network Filesystem, Usenix Summber Conference Proceedings, Summer 1985". Since you mentioned the other Usenix papers, I think you should have also looked at this one (which is where all started). I can't see what is your point wrt the "Stateless Servers" section. The RFC *is* the protocol spec (although it is not 100% complete). The same goes for your comment about the NFSPROC_WRITECACHE procedure. That implied the server's cache (it is mentioned later in RFC-1094), so it is twice irrelevant to this discussion, first because the procedure is not used in NFSv2, and because it has to do with server and not client caching. My point was that the protocol spec does not discuss client caching; therefore client caching is an implementation property. For practical reasons, NFS client implementations do caching. That does *not* mean they violate the protocol. Panos -- Panos Tsirigotis Email: panos.tsirigotis@eng.sun.com Standard disclaimer: I am speaking only for myself, and not for Sun.