Re: [nfsv4] Ben Campbell's No Objection on draft-ietf-nfsv4-rpcrdma-bidirection-07: (with COMMENT)

[Chuck Lever <[e-mail]>]

> On Mar 1, 2017, at 3:23 PM, Ben Campbell <[e-mail]> wrote:
> 
> Ben Campbell has entered the following ballot position for
> draft-ietf-nfsv4-rpcrdma-bidirection-07: No Objection
> 
> When responding, please keep the subject line intact and reply to all
> email addresses included in the To and CC lines. (Feel free to cut this
> introductory paragraph, however.)
> 
> 
> Please refer to https://www.ietf.org/iesg/statement/discuss-criteria.html
> for more information about IESG DISCUSS and COMMENT positions.
> 
> 
> The document, along with other ballot positions, can be found here:
> https://datatracker.ietf.org/doc/draft-ietf-nfsv4-rpcrdma-bidirection/
> 
> 
> 
> ----------------------------------------------------------------------
> COMMENT:
> ----------------------------------------------------------------------
> 
> Some minor comments:
> 
> -4.1: This is the first mention of "credits", and there is no definition.
> I realize that the term is defined in the reference from the previous
> section. It would be helpful to mention that in the context of that
> reference.

I can add that.


> --  Are there any head-of-line-blocking issues introduced by
> bidirectional transactions? For example, can a reply get stuck behind
> requests that are blocked by flow control?

Head-of-line-blocking issues that affect one direction
should have no impact on the other direction because the
two directions have independent credit accounting
mechanisms.


> -5.4, 4th paragraph, last sentence: Can a reverse requestor reasonably
> give up or time out, rather than wait "indefinitely"?

Yes, a timeout is a reasonable response to this situation.
I can soften this sentence to include that possibility.


> -8:  This implies that reverse direction transactions do not change
> anything.If that is the case, please say so explicitly. For example, Is
> there any change to authentication for reverse calls?

Reverse calls would make use of a separate authentication
mechanism and credentials, but would otherwise operate in
the same fashion as requests in the forward direction. I
can add a sentence that states this.


> I am not an expert
> in direct memory access transport protocols; are there every situation
> where authentication depends on an initial request from the client?

The authentication mechanisms are handled by the RPC layer.
There is no authentication (or integrity or confidentiality
protection) in RPC transports, whose only responsibility is
moving bytes. I can add an introductory sentence that states
this.


--
Chuck Lever