NFS is an 'open' standard?
O. Scott Sands
that Sun gets a 'cut' on it's deployment and maintains a
proprietary lock on it.
I know that it is used in all of the freeware OS's and that it's
been published in and RFC. But does this REALLY make it
open? How tight is Sun's grip on NFS?
Anybody?
-- O. Scott Sands o.s....@ieee.org
Ric Werme
>Somebody suggested to me that NFS is not an open standard;
>that Sun gets a 'cut' on it's deployment and maintains a
>proprietary lock on it.
That body is highly misinformed. The final trigger that pushed the
creation of NFS V3 was DEC's, err, Digital's, err, Compaq's release of
OSF/1, err, DIGITAL UNIX, err, Compqa's Tru64 UNIX. The only FS that
did not support > 32 bit FSes was NFS. While V3 support didn't make
it into Tru64 until release 3.0, it was the first release shipped. Sun
didn't get a cut, certainly not for the protocol.
>I know that it is used in all of the freeware OS's and that it's
>been published in and RFC. But does this REALLY make it
>open? How tight is Sun's grip on NFS?
If you join the IETF mail list for NFS V4, you'll see a lot of effort
from Network Appliance, Transmeta, Hummingbird, etc.
>--------------70FC45608182C135229EDCC6
>Content-Type: text/html; charset=us-ascii
>Content-Transfer-Encoding: 7bit
><!doctype html public "-//w3c//dtd html 4.0 transitional//en">
><html>
>Somebody suggested to me that NFS is not an open standard;
Please don't post HTML to _any_ USENET newsgroup. It's very bad form.
--
Ric Werme | http://people.ne.mediaone.net/werme
we...@nospam.mediaone.net | http://www.cyberportal.net/werme
^^^^^^^ delete
Mark A. Crampton
>
> Somebody suggested to me that NFS is not an open standard;
> that Sun gets a 'cut' on it's deployment and maintains a
> proprietary lock on it.
>
> been published in and RFC. But does this REALLY make it
> open? How tight is Sun's grip on NFS?
>
Crap. I wish I knew the answer. I don't think it's a POSIX or whatever
standard; it's more of a "defacto" standard, from what little I know.
Of course, Micro-something lives on this defacto idea.
All reasonable OS's do have NFS (to buy or embedded). Actually, the
more I think about it, there are RFC's for it so it IS a standard!!!!!
There is NFS V2 and V3 and (whatever else). It IS a standard, but it
was "invented" by Sun, albeit other versions of something similar have
been around here and there. USL tried their RFS standard, which was
(loosely) similar.
So, after much blabbing while dealing with jet lag: YES, it is a
standard, so tell your compatriot that it is (and whatever else you want
to say).
Barry Margolin
Mark A. Crampton <crampton._...@indy.net> wrote:
>All reasonable OS's do have NFS (to buy or embedded). Actually, the
>more I think about it, there are RFC's for it so it IS a standard!!!!!
They're "Informational" RFC's (just copies of Sun's documentation), so
they're still just de facto standards.
--
Barry Margolin, bar...@bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.
Rick Macklem
: Somebody suggested to me that NFS is not an open standard;
: that Sun gets a 'cut' on it's deployment and maintains a
: proprietary lock on it.
: I know that it is used in all of the freeware OS's and that it's
: been published in and RFC. But does this REALLY make it
: open? How tight is Sun's grip on NFS?
: Anybody?
Well, here's my vague recollections on it...
Sun introduced NFS Version 2 (Version 1 was just an in-house development
and was never turned loose on the world) in about 1986. It published the
specification of the protocol in one of the Sun OS manuals that came with
Sun OS, but could be purchased separately without signing and non-disclosure
agreement. They also provided a "reference port" under non-disclosure that
most vendors used as the basis of their NFS implementation. (My understanding
was that vendors either paid a royalty to Sun per copy shipped or a one time
lump sum that allowed them to ship unlimited copies. I don't know which
vendors did which?)
In the late eighties, I believe that my implementation was the first full
client/server implementation based on the specification, but without access
to the reference port. (Since UC Berkeley had a Sun reference port license
and sources, Sun was at one point concerned that their code was being used.
Fortunately, the University of Guelph did not have a reference port license
and I had never seen the Sun code, so I was able to work in a "clean room"
environment.) Although the specification did tell you where all the bits
and bytes go, it did not say much about semantics of the RPCs. The latter
had to be gleaned from published papaers, such as the Sun Usenix one on
NFS and interoperability testing between machines running NFS from Sun.
The spec. was re-published as an RFC in 1987, but the protocol was
unchanged from the spec. in the Sun manual that was a part of the
Sun OS (Version 3?) documentation. However, the associated locking
protocol was not published and, as such, was at that point still not
freely available. A bare bones spec. of the locking protocol was
published as a part of the X/Open document on NFS, that was freely
available (well, actually it cost about $75 bucks, but you didn't have
to sign anything:-).
By the time Version 3 came along, Sun was still the "curator" of the
protocol, but it seems that it was a pretty open industry consultative
process that defined the protocol (I wasn't involved in this, but that
was my impression). I do know that the V3 protocol was published,
including drafts available during its development. I don't know if Sun
licensed a reference port for V3 and I don't know if any other vendor's
implementations were based upon Sun code for V3.
Finally, as V4 is being developed, Sun has handed the protocol over to
IETF, which is now the "curator" of it. (I can't remember the correct
term for this, but you can look on the IETF ftp site for information.)
In summary, it depends upon your definition of Open. Although most
vendors used the Sun reference port and paid (still pay?) royalties
to Sun, Sun has never claimed the protocol as proprietary and has not
tried to stop freely available implementations from being built (in fact
they have gone out of their way to help developers of freely available
implementations, like myself, get the protocol right and avoid interop.
problems). IMHO, they have been "good curators" of the protocol over the
years, and I am actually a bit concerned that the new IETF mechanism
will become more onerous. At this point, the NFS V4 draft looks pretty
good, but it is already a lot more complex than its predecessors and
still growing. Compare this to the AFS situation, where a good protocol
has been stifled by its proprietary nature (flame away on this one, if
you want:-).
Does this help clear up the murky waters? rick
Paul Eggert
>They're "Informational" RFC's (just copies of Sun's documentation), so
>they're still just de facto standards.
There are standards and there are standards. As far as the ISO is
concerned, Internet RFCs are not standards. An ISO standard cannot
refer to an Internet RFC, except in non-normative commentary (where it
can refer to anything it wants to, even Gilbert&Sullivan say).
So, in the official and bureaucratic sense of the term, yes, NFS is not
standardized. But TCP/IP is not standardized either. :-)
Ric Werme
>Barry Margolin <bar...@bbnplanet.com> writes:
>>They're "Informational" RFC's (just copies of Sun's documentation), so
>>they're still just de facto standards.
>There are standards and there are standards.
The holder of the Unix trademark, the Open Group, promulgates both
standards and branding suites. Check
http://www.opengroup.org/pubs/catalog/c702.htm:
Protocols for Interworking: XNFS, Version 3W
Open Group Technical Standard
C702 ISBN 1-85912-184-5 February 1998 352 pages.
There is no Corrigendum for this Document
Was C525 Protocols for X/Open Interworking: XNFS, Issue
4 Version 3 8/96
INTERWORKING: A complete solution to transparent file
access aligned with Sun's NFS Version 3 and its
WebNFS(TM) extensions.
ABSTRACT: This Open Systems NFS (XNFSv3W) Open
Group Technical Standard is aligned with Sun's NFS
Version 3, and incorporates the Sun WebNFS(TM)
extensions. The process of accessing remote files and
directories as though they were part of the local file system
hierarchy is commonly known as 'transparent file access'
(TFA). The most widely used heterogeneous TFS
architecture is the Network File System (NFS), originally
developed by Sun Microsytems. The Open Group's XNFS
offers a complete solution to transparent file access between
'open-system-compliant' systems, through the XNFS
protocols for interoperability, and The Open Group's XSI
interfaces (defined in Document C606, see Document Set
T912) for application/user portability (as identified in several
XNFS appendixes).
My copy is at the office, but I think it covers NFS V2, V3 (with WebNFS)
STAT, NLM, and MOUNT. One might point out the reference to Sun's NFS V3,
but as Sun did the majority of the work on the V2 and V3 speacs and
worked with the Open Group to produce the standard, I don't complain.
Several companies influenced NFS V3 development.
>So, in the official and bureaucratic sense of the term, yes, NFS is not
>standardized. But TCP/IP is not standardized either. :-)
Having gone through the the XNFS branding process (with the beta test
branding suite), I can assure you it adequately enough standardized to
cause companies of many sizes to jump through a lot of hoops, some with
sharp edges.
I'd have to ask the networking folks about TCP, so you may want to look
that up yourself.
NFS V4 is being developed in a IETF working group. It will be an IETF
standard. Anyone can join in.
-Ric Werme
Barry Margolin
Paul Eggert <egg...@twinsun.com> wrote:
>There are standards and there are standards. As far as the ISO is
>concerned, Internet RFCs are not standards. An ISO standard cannot
>refer to an Internet RFC, except in non-normative commentary (where it
>can refer to anything it wants to, even Gilbert&Sullivan say).
So? The ISO is not the final arbiter of standards, the organizations that
develop and purchase products are. The only networking standards the ISO
has produced are the OSI standards, and few of them have become successful
products, so those standards are practically worthless. Organizations that
purchase networking products ask their vendors for products that conform to
TCP/IP, by which they mean the standards that the IETF has approved. IETF
is as good a standards body as any other, it just doesn't happen to use
ISO-approved processes.
Paul Eggert
>Paul Eggert <egg...@twinsun.com> wrote:
>>Barry Margolin <bar...@bbnplanet.com> writes:
>>>They're "Informational" RFC's (just copies of Sun's documentation), so
>>>they're still just de facto standards.
>>There are standards and there are standards. As far as the ISO is
>>concerned, Internet RFCs are not standards.
>So? The ISO is not the final arbiter of standards, the organizations that
>develop and purchase products are.
I quite agree, and made the same point in comp.std.c recently when
someone incorrectly claimed that C == ISO C. But I was responding to
your remark that the NFS standards are still ``just'' de facto.
If the final arbiter of standards are the people that develop and
purchase products, then that ``just'' was out of place,
as de facto computer standards are as important as de jure ones.
The ISO is pretty much in charge of de jure standards.
From the ISO's point of view, the IETF's standards, including those for
TCP/IP, are ``just'' de facto standards. Bizarre, isn't it?
Barry Margolin
What about ANSI (and its counterparts in other countries) and IEEE?
>From the ISO's point of view, the IETF's standards, including those for
>TCP/IP, are ``just'' de facto standards. Bizarre, isn't it?
The basic TCP/IP specifications are also FIPS and MIL-STD standards, which
makes them official standards for the US government and military.
Eric Werme - replace nospam with werme
>Barry Margolin <bar...@bbnplanet.com> writes:
>>So? The ISO is not the final arbiter of standards, the organizations that
>>develop and purchase products are.
>I quite agree, and made the same point in comp.std.c recently when
>someone incorrectly claimed that C == ISO C. But I was responding to
>your remark that the NFS standards are still ``just'' de facto.
>If the final arbiter of standards are the people that develop and
>purchase products, then that ``just'' was out of place,
>as de facto computer standards are as important as de jure ones.
In that spirit, one can argue that a vendor's NFS should be viewed
suspiciously until they've been to Connectathon and back. In fact,
I have a lot more respect for vendors who do that than implement to
the spec and declare victory. Fortunately, just about all the
vendors of note make it to Connectathon. Every year we find some
sort of interoperability problem in our code or someone elses that
I'm relieved to have found in that setting.
http://www.connectathon.org/
--
<> Eric (Ric) Werme <> The above is unlikely to contain <>
<> ROT-13 addresses: <> official claims or policies of <>
<> <jr...@mx3.qrp.pbz> <> Compaq Computer Corp. <>
<> <jr...@zrqvnbar.arg> <> http://people.ne.mediaone.net/werme <>
Paul Eggert
>In article <7ooefn$l33$1...@shade.twinsun.com>,
>Paul Eggert <egg...@twinsun.com> wrote:
>>The ISO is pretty much in charge of de jure standards.
>What about ANSI (and its counterparts in other countries) and IEEE?
I think both ANSI and IEEE defer to the ISO, but I don't pretend to
know all the ins and outs of their relationships.
I've heard that the IEEE does allow normative references to IETF
standards, which would make POSIX an interesting case, since it's
developed under IEEE auspices but is an ISO standard. (Presumably the
ISO editors go through and turn any normative references to IETF
documents into non-normative references. :-)
I don't know whether ANSI allows normative references to IETF standards.
O. Scott Sands
> O. Scott Sands (san...@home.com) wrote:
>
> : Somebody suggested to me that NFS is not an open standard;
> : that Sun gets a 'cut' on it's deployment and maintains a
> : proprietary lock on it.
>
> : I know that it is used in all of the freeware OS's and that it's
> : been published in and RFC. But does this REALLY make it
> : open? How tight is Sun's grip on NFS?
>
> : Anybody?
>
> Well, here's my vague recollections on it...
>
> Sun introduced NFS Version 2 (Version 1 was just an in-house development
[stuff deleted]
Thank you for your extensive recount of the events pertaining
to the development of NFS. I guess that my question relys to
heavily on the defintion of 'open' which is not really defined.
However, I would say that, if NFS isn't open then nothing is.
My question was prompted by the suggestion that somebody made
that SMB is open and NFS is not. He has long since cried
'uncle' and so my original object have has been met several times
This whole
RL 'Bob' Morgan
RFC 2203 specifies the use of GSS-API, and hence any underlying security
mechanisms that can use GSS-API, with ONC RPC, and hence NFS. I believe
that Sun's NFS implementation in later Solarises (>= 2.6?) supports this,
so that one can authenticate a NFS operation using Kerberos 5, and
encrypt the NFS data stream too if desired. I'm curious to know if there
are interoperable implementations available for any other OS platforms.
Looking thru AIX docs it looks like IBM supports the AUTH_DH
(Diffie-Hellman exchange of DES keys) flavor, but not GSS.
Thanks,
- RL "Bob" Morgan
UW C&C