[GFS-WG] Fw: RNS 1.1 and misc other documents

[Mark Morgan]

Regarding the corrected XML schema and wsdl:  I'm not at all
surprised that there were some small problems along those lines.
Thank you for fixing them.  As near as I can tell without doing
an implementation, everything looks good in the revised documents.

I am confused by your second point.  RNS 1.1 is like a hash map
in a programming language like C where the implementation maps
string names to void*'s.  The map doesn't know how to delete the
data items when a map entry is removed, but that doesn't mean that
the entry can't be removed.  Yes, you are right that there is no
way to know when the entry isn't referenced any more, but that's no
different then the case with a sym link in Unix.  We have tried to
do reference counting in the past with Legion and its a major
nightmare because of failure modes and security (who has permission
to update link counts).  In our opinion, and this is largely
supported by experience with Genesis II, it's preferrable to assume
that all entries are not reference counted.

-Mark

On Mon, 2009-02-02 at 16:06 +0900, Osamu Tatebe wrote:
> Hi All,
> 
> We have some experience with RNS 1.1 and found these issues.
> 
> 1. WSDLs in the spec (RNS and WS-Iterator) include some mistakes.
>    Corrected XML Schema and WSDL are attached.
> 
> 2. There is no way to know an entry is not referrenced to any
>    more.  Regarding RNS entry, life time seems not to be
>    appropriate since users do not assume it will be expired.
>    This means there is no way to delete RNS entries.
> 
> The item 1 is obvious.  How do you think about the item 2?
> 
> Also, we are ready for RNS interoperability test.  If some
> other party has the RNS 1.1 implementation please let us know.
> 
> Thanks,
> Osamu
> 
> On Tue, 13 Jan 2009 10:45:32 +0100
> Erwin Laure <erwin.laure at cern.ch> wrote:
> 
> > Hi,
> > 
> > I haven't seen any further comments to this. As I believe the points 
> > raised below can be addressed quite easily and don't block pushing the 
> > spec out for public comments.
> > 
> > So I call the group for final comments before pushing the RNS 1.1 spec 
> > out for public comments by January 20.
> > 
> > Cheers,
> > 
> > -- Erwin
> > 
> > Osamu Tatebe wrote:
> > > Hi all,
> > > 
> > > After the publication of RNS Specification (GFD.101), we have
> > > discussed about interoperability and ambiguous issues in the
> > > spec.  Attached there are four docs that Andrew's group
> > > prepared including RNS 1.1 proposed spec.
> > > 
> > > The original agreement was OGSA Naming WG would take over after
> > > 1.0, but by the following discussion with area directors and
> > > Andrew's group, GFS-WG proceeds with the discussion of RNS Spec.
> > > 
> > > Regarding the WS-Iterator Spec, the home WG is not decided yet,
> > > but we would like to discuss it since we really need it.
> > > 
> > > Please feel free to provide comments and questions about these
> > > specs.  We would like to finalize the discussion before the
> > > Chrismas time.
> > > 
> > > Here are comments from our group.
> > > 
> > > - Multiple endpoints at each entry are requested by many parties
> > >   mostly because of management of file replicas.  I know it would
> > >   be managed by WS-Naming, but allowing multiple endpoints provides
> > >   much more flexibility.  Also, even when WS-Naming is not supported,
> > >   we can manage file replicas.
> > > 
> > > - By performance reason, we would like to consider operations to add
> > >   multiple entries at once.
> > > 
> > > - Lookup operation now accepts exact entry name or NULL that means
> > >   directory entry listing.  Is specifying wildcard to filter entries
> > >   useful?
> > > 
> > > Thanks,
> > > Osamu
> > > 
> > > Begin forwarded message:
> > > 
> > > Date: Thu, 17 Jul 2008 11:43:58 -0400
> > > From: "Andrew Grimshaw" <grimshaw at virginia.edu>
> > > To: "'David E Martin'" <martinde at us.ibm.com>, "'Erwin Laure'" <erwin.laure at cern.ch>, "'Osamu Tatebe'" <tatebe at cs.tsukuba.ac.jp>
> > > Cc: "'Mark Morgan'" <mmm2a at virginia.edu>, "'Karolina Sarnowska'" <kas9ud at virginia.edu>
> > > Subject: RNS 1.1 and misc other documents
> > > 
> > > 
> > > All,
> > > 
> > > Here are the documents we promised - they are of course drafts. There are
> > > four docs.
> > > 
> > >  
> > > 
> > > 1)    The RNS 1.1 proposed spec - see appendix A on what is different from
> > > the 1.0
> > > 
> > > 2)    WS-Iterator draft
> > > 
> > > 3)    A white paper on why we chose not to use WS-Enumeration
> > > 
> > > 4)    The RNS 1.1 OGSA WSRF Basic Profile Rendering 1.0
> > > 
> > >  
> > > 
> > > Enjoy
> > > 
> > >  
> > > 
> > > A
> >