RE: draft-11 EXCHANGE_ID

On Thu, 21 Jun 2007, Noveck, Dave wrote:

>> 1199: client ID "is changed whenever the
client or the server
> reinitializes"
>>       Not true, perhaps something like "may be
changed..."
>
> Why is this not true?  Not sure what you are getting at
here.
>

The client can reboot and change it's verifier w/o calling
EXCHANGE_ID to 
get a new client ID.

The server can reboot and retain enough info that the client
is not forced 
to call EXCHANGE_ID and get a new client ID.

I read "is changed *whenever* the client or the server
reinitializes" to 
imply that in the above situations a new client ID must be
generated, 
which is not true.

 	Fred


_______________________________________________
nfsv4 mailing list
nfsv4ietf.org
https://
www1.ietf.org/mailman/listinfo/nfsv4

--- Fredric Isaman <iisamanciti.umich.edu> wrote:

> 
> 
> On Thu, 21 Jun 2007, Noveck, Dave wrote:
> 
> >> 1199: client ID "is changed whenever the
client or the server
> > reinitializes"
> >>       Not true, perhaps something like
"may be changed..."
> >
> > Why is this not true?  Not sure what you are
getting at here.
> >
> 
> The client can reboot and change it's verifier w/o
calling
> EXCHANGE_ID to 
> get a new client ID.

How can it change its verifier without calling EXCHANGE_ID?

> 
> The server can reboot and retain enough info that the
client is not
> forced 
> to call EXCHANGE_ID and get a new client ID.
> 
> I read "is changed *whenever* the client or the
server reinitializes"
> to 
> imply that in the above situations a new client ID must
be generated,
> 
> which is not true.

I agree that server reinitialization, for some definitions
of
reinitialization, does not imply the client ID is tossed.
I will attempt to clarify this in draft-12. I'll also go
over
your other comments.
> 
>  	Fred
> 
> 
> _______________________________________________
> nfsv4 mailing list
> nfsv4ietf.org
> https://
www1.ietf.org/mailman/listinfo/nfsv4
> 


_______________________________________________
nfsv4 mailing list
nfsv4ietf.org
https://
www1.ietf.org/mailman/listinfo/nfsv4

> The server can reboot and retain enough info that the
client is not
forced 
> to call EXCHANGE_ID and get a new client ID.

Sure but this is an introductory section.  Trying to deal
with
all those possibilities is going to make it hard to convey
the
concepts to the user.  After all, we are saying "is
changed"
and not "MUST be changed".

In particular, saying "may be changed" is going to
give the
reader a completely wrong idea, i.e. that this is in some
way
optional, in that you can simply decide not to do it.

The fact that you can save enough state so that you can
reboot
and still not reinitialize, from the v4 protocol point of
view, 
is a subtlety that may be worth mentioning somewhere but if
we do it here, I'm afraid it might make someone's head
explode.



-----Original Message-----
From: Fredric Isaman [mailto:iisamanciti.umich.edu] 
Sent: Thursday, June 21, 2007 12:32 PM
To: Noveck, Dave
Cc: nfsv4ietf.org
Subject: RE: [nfsv4] draft-11 EXCHANGE_ID




On Thu, 21 Jun 2007, Noveck, Dave wrote:

>> 1199: client ID "is changed whenever the
client or the server
> reinitializes"
>>       Not true, perhaps something like "may be
changed..."
>
> Why is this not true?  Not sure what you are getting at
here.
>

The client can reboot and change it's verifier w/o calling
EXCHANGE_ID
to 
get a new client ID.

The server can reboot and retain enough info that the client
is not
forced 
to call EXCHANGE_ID and get a new client ID.

I read "is changed *whenever* the client or the server
reinitializes" to

imply that in the above situations a new client ID must be
generated, 
which is not true.

 	Fred

_______________________________________________
nfsv4 mailing list
nfsv4ietf.org
https://
www1.ietf.org/mailman/listinfo/nfsv4