Re: [nfsv4] IETF 102 nfsv4 WG meeting go/no-go

[Tom Haynes <[e-mail]>]

> On May 17, 2018, at 2:46 PM, J. Bruce Fields <[e-mail]> wrote:
> 
> On Wed, May 16, 2018 at 08:22:26PM -0700, Thomas Haynes wrote:
>>> On May 16, 2018, at 8:20 PM, Dr James Bruce Fields <[e-mail]> wrote:
>>> On Wed, May 16, 2018 at 11:10:54PM -0400, Trond Myklebust wrote:
>>>> On Wed, 16 May 2018 at 21:26, J. Bruce Fields <[e-mail]> wrote:
>>>>>> draft-haynes-nfsv4-delstid-00 <
>>>> https://datatracker.ietf.org/doc/draft-haynes-nfsv4-delstid/>
>>>> 
>>>>> What's the motivation for WANT_OPEN_XOR_DELEGATION?
>>>> 
>>>> 
>>>> What's the motivation for asking for a delegation _and_ an open stateid?
>>>> That means you have to return 2 objects instead of just one.
>>>> 
>>>> The WANT_OPEN_XOR_DELEGATION  option authorises the server to return a
>>>> delegation without the open stateid, which cuts down on the amount of
>>>> traffic between the client and server. Yes, you could just ask for a
>>>> delegation using WANT_DELEGATION, but if the server is unable to hand one
>>>> out, it will return an error, and you're back to sending an OPEN..
>>> 
>>> Got it, thanks.  So the traffic saved is one unnecessary CLOSE (in the
>>> case the server returns a delegation)?
>>> 
>>> --b.
>> 
>> Yes - and perhaps a GETATTR in that compound
> 
> Is there a particular case you've encountered when those turn out to be
> a problem?
> 
> --b.

No, but in general we like to reduce the number of GETATTRs, especially ones
which require us to go to the DSes.