Re: [kitten] [nfsv4] draft-ietf-nfsv4-rpcsec-gssv3: request for review

["Adamson, Andy" <[e-mail]>]

On Aug 4, 2014, at 4:01 PM, Benjamin Kaduk <[e-mail]> wrote:

> On Mon, 4 Aug 2014, Nico Williams wrote:
> 
>> On Mon, Aug 04, 2014 at 04:55:57PM +0000, Adamson, Andy wrote:
>>> On Aug 1, 2014, at 6:15 PM, Nico Williams <[e-mail]> wrote:
>>>> (Was "multi-principal" my name for this?  No, I called them compound
>>>> authentication.  I prefer "compound”.)
>>> 
>>> Hi NIco
>>> 
>>> I changed the name from ‘compound’ to multi-principal’ in response
>>> the review comments at IETF 89 where many NFSv4 WG members expressed
>>> that ‘compound’ had too many meanings (especially in NFSv4.x) and
>>> led to confusion.
>> 
>> "Compound" is used as an adjective in all cases.  I don't see how
>> "compound authentication" (or "compound context handle", ...) is
>> confusable with "compound RPC".  But this isn't important enough to me;
>> aligning the terminology with AFS' rxgk is.  Ben, what does rxgk call
>> this?
> 
> If I remember correclty from the talk from Toronto, Andy was not quite so keen to align with rxgk, but that may have been based on incomplete data.

Well, I looked at the AFS ‘combined’ tokes, and thought they were close, but not close enough to re-use the name.

—>Andy
> 
> Anyway, for rxgk, we don't talk very much about the actual ~compound token; the operation itself is CombineTokens or AFSCombineTokens, with the latter being the operation which is analogous to the rpcsec case.  The AFSCombineTokens case is a little convoluted, since rxgk-afs differentiates between "vlserver tokens" (roughly analogous to the MDS) and "fileserver tokens" (analogous to the DS).  AFSCombineTokens takes vlserver tokens as input and produces fileserver tokens as output, so a lot of the time they are just referred to as "fileserver tokens".
> 
> There are a couple places where we refer to the CombineTokens output as a "combined token", which is I think what you're looking for as an answer. That doesn't look like it translates very nicely to the rpcsec case, though. :(
> 
> -Ben