Re: [nfsv4] RFC 8276 and minor versions

Hi Frank,

thanks for bringing this up.

----- Original Message -----
> From: "Dave Noveck" <davenoveck@gmail.com>
> To: "van der Linden, Frank" <fllinden=40amazon.com@dmarc.ietf.org>
> Cc: "NFSv4" <nfsv4@ietf.org>
> Sent: Sunday, March 29, 2020 10:09:02 PM
> Subject: Re: [nfsv4] RFC 8276 and minor versions

> On Fri, Mar 27, 2020 at 1:05 PM van der Linden, Frank <fllinden=
> 40amazon.com@dmarc.ietf.org> wrote:
> 
>> Hi everyone,
>>
>> Quick question:
>>
>> It's my reading of RFC 8276 and RFC 8178 that 8276 (user extended
>> attributes) extends the (then) current minor version - 4.2.
> 
> 
> Yes.
> 
> 
>> So 4.1 or 4.0 servers/clients can not use this functionality.
>>
> 
> A server or client that did that would not using or implementing 4.0 or
> 4.1.  It would be implementing its own private protocol.
> 
> Is that correct?
> 
> 
> It certainly was the intention when RFC8178 was written.
> 
> I'm not sure that the right decision was made but realistically, the
> concept of extendible versions was so new that it would have been hard to
> declare an existing minor versions extensible post facto.
> 
> The wording doesn't seem to be that clear,
> 
> 
> I think it is clear enough.  RFCs 5661 and 7530 are presented as fixed,
> with fixed sets of operations and fixed (in RFCs 5662 and 7531)  XDR files
> that defined them.   There is nothing in RFC8178 that changes that or calls
> it into question.
> 
> but I might have missed something, so hence my question.
>>
> 
> You didn't miss anything but it appears I did.   At the time, extensible
> versions were considered so weird/dubious that I never thought to
> clearly state that existing minor versions could not be extended.  At the
> time it seemed too *obvious *to state explicitly.
> 
> This came up because there's no technical issue with implementing 8276 on
>> top of 4.1 or even 4.0.
> 
> 
> True.
> 
> 
>> But my reading is that this would violate the RFCs.
>>
>> True, but RFCs can be changed if the working group feels it is worth the
> effort.
> 
> For v4.0, it woyld be a substantial effort which probably isn't worth it.
> 
> For v4.1, it is not worth it for a complementary reason.   It is easy to
> take a v4.1 server and turn it into a v4.2 server that does not support any
> of the v4.2 OPTIONAL extensions.   It just has to recognize them, parse the
> ops, and return NFS4ERR_NOTSUPP if the minor version is 2.   At that point
> it has become a v4.2 server and the ops within rfc8276 can be
> implemented within it in accord with RFC8178.

To me it's just looks weird for a server to claim to support 4.2 and respond with
NFS4ERR_NOTSUPP to all operations just to add a support for capability that is
independent of any 4.2 functionality. However, in did not a big deal.

Tigran.

> 
> Thanks,
>>
>> - Frank
>>
>> _______________________________________________
>> nfsv4 mailing list
>> nfsv4@ietf.org
>> https://www.ietf.org/mailman/listinfo/nfsv4
>>
> 
> _______________________________________________
> nfsv4 mailing list
> nfsv4@ietf.org
> https://www.ietf.org/mailman/listinfo/nfsv4