Re: [nfsv4] Global namespace for v4.1?

brent@eng.sun.com

Message-ID: <3FDE0231.7060108@eng.sun.com>
From: Brent Callaghan <brent@eng.sun.com>
Subject: Re: [nfsv4] Global namespace for v4.1?
Date: Mon, 15 Dec 2003 10:49:21 -0800

Wouldn't it be better to have a single method of accessing
name-to-location mappings that can be used by both clients
and servers ?

I'm having trouble reconciling "dumb" with "NFSv4 client." :-)

	Brent

Noveck, Dave wrote:
> And given that you are going to support such dumb clients, what is
> the value of creating additional mechanisms to support smart clients?
> There's an existing mechanism which seems to meet the need whereby
> clients get this information, from the server by means of a referral,
> (and it is a mechanism that clients need anyway to deal with migration),
> and having them use that, no matter now smart they are, seems the most 
> sensible path to take.  
> 
> -----Original Message-----
> From: Nicolas Williams [mailto:Nicolas.Williams@sun.com]
> Sent: Monday, December 15, 2003 1:18 PM
> To: Brent.Callaghan@sun.com
> Cc: nfsv4@ietf.org
> Subject: Re: [nfsv4] Global namespace for v4.1?
> 
> 
> On Mon, Dec 15, 2003 at 10:12:13AM -0800, Brent Callaghan wrote:
> 
>>Nicolas Williams wrote:
>>
>>>IMO we need:
>>>
>>>- a referral mechanism (got that)
>>>- a "root" finding mechanism (we can build that today as a set of
>>>  conventions, really)
>>
>>Yes, a couple of important components, though, I should point out,
>>these elements are just a few in a long list of the list of features
>>that we expect of any general-purpose name service, e.g. LDAP, NIS, DNS etc.
> 
> 
> I acknowledged as much, and should have put that on my list.
> 
> 
>>I think what Julian is trying to point out is that a namespace is more
>>than just a collection of exported filesystems.  There's some expectation
>>that these filesystems will be organized in some way, most likely one or
>>more hierarchical trees, and their shape will be determined by the customer.
>>
>>I predict that as the requirements of the namespace become more clearly
>>defined, the advantages of using an established name service will become
>>more apparent.
> 
> 
> Surely.  But how the information is distributed to either smart clients
> or smart servers is a problem that I see as separate from the question
> of whether or not we should support dumb clients (for which we need
> referrals).
> 
> Nico

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