Re: comments on draft-shepler-nfsv4-02-01.html

New Message Reply About this list Date view Thread view Subject view Author view Attachment view

From: Mike Eisler (mre@eng.sun.com)
Date: 02/22/99-07:34:43 PM Z


Date: Mon, 22 Feb 1999 17:34:43 -0800 (PST)
From: Mike Eisler <mre@eng.sun.com>
Subject: Re: comments on draft-shepler-nfsv4-02-01.html
Message-ID: <Roam.SIMC.2.0.6.919733683.4813.mre@eng.sun.com>

> > >    Name:           read_max
> 
> > 	In the event this attribute doesn't exist, I'd simply try
> > 	a read of size X, and either get X bytes, or get a short read.
> > 
> > 	Either make it mandatory or get rid of it, but as RECOMMENDED,
> > 	I consider it it of no use.
> 
> > >    Name:           write_max
> 
> > I don't think we need to make this MANDATORY because obviously
> > if a server say supports only read only file systems, WRITE is
> > nonsequitur. But I'd like to see some language that says that if
> > the server supports WRITEs on the object, it is strongly RECOMMENDED
> > that it support this attribute, because otherwise the client
> > may guess too conservatively or too liberally, and the result
> > is that performance suffers and/or bandwidth is wasted.
> 
> The benefits of having this are clear, as you say.  The only argument
> the other way is about whether it is really essential for a minimalist
> server to implement this to be useful.

True. Ok, I withdraw my comment about read_max. As for write_max I think
we are in agreement, but I just want the language on the attribute to
be strong as I described above.

> > >    Name:           fsid
> 
> > Craig Everhart said several months ago that the case can be
> > made that the fsid should be a tuple, at least a two-tuple.
> > I agree. The DCE/DFS local file system needs it, and it would
> > also ease implementation of file sets. So I think fsid should be
> > split into fsid.major and fsid.minor, each of type uint64.
> 
> This sounds good to me.
> 
> > Given that, the attributes that apply to a "filesystem" ought to
> > have counter parts that apply to an fsid.minor as well.
> 
> Could you please expand on this?  Would we want to duplicate all of
> those types of fields?  Or would we want to sort somehow?  I would
> think the client would want to find out the limits (e.g. free space)
> and characteristics of the directory under operation, and not care
> to think hard about two sets of parameters.

"free space" is an excellent example of an attribute that could have a
different value for each fsid.minor within the same fsid.major.

If the client is aware that two different file objects have a same
fsid.major but different fsid.minor, then the client will want to
know the free space on the different fsid.minor's. Say the application
is copying a file to multiple directories, and each directory has
different fsid.minor. 

	-mre


New Message Reply About this list Date view Thread view Subject view Author view Attachment view

This archive was generated by hypermail 2.1.2 : 03/04/05-01:46:41 AM Z CST