Re: [nfsv4] rfc5666bis and rpcrdma-bidirection progress update

[Chuck Lever <[e-mail]>]

> On Nov 28, 2016, at 1:12 PM, Spencer Shepler <[e-mail]> wrote:
> 
>  
> My comments were general, Dave.  Yes, I am the bottleneck on some of this but I get one nag email every few weeks.  Mmm, yes, it has slipped on my todo list but I also have not been reminded with any great frequency – is it urgent or not.  
>  
> Yes, I can improve my cycle time but there are many other steps involved and just waving that it is “IETF Process” is unsatisfactory from my point of view.  It is what you make of it. The elapsed time can be reduced – unfortunately, much of the slowness is self-induced (working group self-induced) and indicative of the relative priorities of everyone involved.

I was informed several months ago that there were at least five I-Ds in the
WG pipeline ahead of rfc5666bis and rpcrdma-bidirection. Three of them have
recently become RFCs, leaving at least two more. The remaining documents I'm
aware of are the multi-domain I-D and the SCSI layout I-D.

So far I haven't felt it necessary to send frequent reminders about my I-Ds
because chair write-ups are not really what's blocking the progress of these
documents. I didn't feel it necessary to nag someone who couldn't do much
about the delays with RPCSEC GSSv3, for example.

It did leave me with a sense of helplessness about this backlog. And I can
admit to having spent less time recently on IETF work because it didn't
seem valuable to pursue, given the existing blockages. Lack of nagging was
not related to relative priorities.

It would help to get regular status updates on the document pipeline,
whether there are any hold-ups, and what the non-executive members of the
WG can do about them. IESG review processes are managed off the WG mailing
list, thus many of us are not privy to progress on individual I-Ds, nor do
we have a sense of what tasks there are remaining to do. The chairs are
usually copied on all of this. Perhaps they could monitor and periodically
reflect a summary to the WG.


> Spencer
>  
> From: David Noveck [mailto:[e-mail]] 
> Sent: Monday, November 28, 2016 9:33 AM
> To: Spencer Shepler <[e-mail]>
> Cc: Chuck Lever <[e-mail]>; [e-mail]
> Subject: Re: [nfsv4] rfc5666bis and rpcrdma-bidirection progress update
>  
> With regard to these documents I suppose that it is appropriate to remind you of the history here:
> 	• These documents have been in the state WG Consensus:Waiting for Writeup for over four months, since 7/19.
> 	• Chuck tried to reduce the workload by deciding to drop the associated informational document.
> 	• Chuck made efforts to remind you of the need to provide these writeups.
> 	• This appeared to culminate in your statement  on 10/25: " For the I-Ds that are "waiting for write-up", those are on my to-do list and will be out in the next week"
> 	• It is now about five weeks since that statement and we have heard nothing about progress in this regard.  
> > Reminders, regularly scheduled conference calls, changing ownership, breaking the tasks into smaller items, discarding unneeded functionality to move the process forward.
>  
> We've tried reminders and Chuck did remove unnecessary work.  However, the IETF rules don't appear to allow changing ownership of the task of producing a wrIteup.  If you think that is appropriate, then perhaps you should investigate it.  Given that there is a bunch of other documents that will soon be in the same state, maybe you should consider that.  Would regularly scheduled conference help, or would they just take up time?
>  
> > most of it is under the control of the authors and the collective working group once there is consensus that an I-D is a WG draft.
>  
> True, but there is a significant portion of the process where we wind up waiting for action by the WG chair, and for these documents, that has been a considerable portion of the delay.  Is there a viable way to spread out this work better?
>  
>  
>  
> On Mon, Nov 28, 2016 at 11:53 AM, Spencer Shepler <[e-mail]> wrote:
>  
> For this comment, “If we want to improve the process, and I believe we should, we need to focus on review quality/adequacy as well as the pace of the process.”
>  
> If there is a desire for an item to move more quickly, treat it like any other item in life – take responsibility for it, invest the time, get others that are needed to do the same.
>  
> There are a variety of techniques for managing work that is done by others.  Reminders, regularly scheduled conference calls, changing ownership, breaking the tasks into smaller items, discarding unneeded functionality to move the process forward.
>  
> The “process” can move more quickly if desired – at least for the items that are under direct control and in the case of Internet Drafts – most of it is under the control of the authors and the collective working group once there is consensus that an I-D is a WG draft.
>  
> Spencer
>  
> From: nfsv4 [mailto:[e-mail]] On Behalf Of David Noveck
> Sent: Sunday, November 27, 2016 3:55 AM
> To: Chuck Lever <[e-mail]>
> Cc: [e-mail]
> Subject: Re: [nfsv4] rfc5666bis and rpcrdma-bidirection progress update
>  
> So the review process will not be interrupted.  It will continue at the same pace as before.  
> 
> With regard to the characterization of the process as "excruciatingly sluggish", it does seem that not very much has happened since the working group reached consensus on  this document on 4/17/2016 (according to the data tracker history).
> 
> However, to be fair, we should note that draft-ietf-nfsv4-rfc5666bis-00 was submitted on 12/01/2015 and it it took the working group less than a year to get to the point we are.    It looks likely to me that the whole process will (cross your fingers) be completed in around 16 months from first submission.   It is hard to get documents done faster than that within the IETF process, which is not optimized for speed.
> 
> Contrast that with RFC5666 itself which took over five years from -00 submission to RFC publication.  Despite the slow pace of the process, it appears that the document did not get the kind of review it needed before publication, making rfc5666bis necessary. I'd like to thank the authors for persevering in the effort to provide NFS with an RDMA-based transport.
> 
> If we want to improve the process, and I believe we should, we need to focus on review quality/adequacy as well as the pace of the process.
> 

--
Chuck Lever