SORT BY:

LIST ORDER
THREAD
AUTHOR
SUBJECT


SEARCH

IPS HOME


    [Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

    iSCSI version numbers



    Let me also point out that Matt's expectation of
    interoperability of implementations based on
    Internet-Drafts is misplaced.  Vendors who ship
    products based on I-Ds do so at their own risk -
    there's more than enough warning that things are
    subject to change in the standard Internet-Draft
    boilerplate.
    
    My preference would be to hold the version number
    at 0 until the iSCSI spec is done.  This will give
    vendors who have shipped to Internet-Draft versions
    an incentive to promptly update their code.
    If future plugfest events need to test multiple
    implementation variants, they can do roughly what
    was done this time - temporarily define the version
    numbers that will distinguish the variants *for that
    event only*.  I would suggest use of *large* numbers
    for this purpose to avoid this sort of confusion
    in the future.
    
    Thanks
    --David
    
    ---------------------------------------------------
    David L. Black, Senior Technologist
    EMC Corporation, 42 South St., Hopkinton, MA  01748
    +1 (508) 435-1000 x75140     FAX: +1 (508) 497-8500
    black_david@emc.com       Mobile: +1 (978) 394-7754
    ---------------------------------------------------
    
    
    > -----Original Message-----
    > From:	Robert D. Russell [SMTP:rdr@mars.iol.unh.edu]
    > Sent:	Wednesday, July 18, 2001 8:23 PM
    > To:	Matt Wakeley
    > Cc:	ips
    > Subject:	Re: London: Call for agenda items
    > 
    > Matt:
    > 
    > The problem is that there is no consensus version 1.
    > Most current "draft 6" implementations are really implementing
    > "draft 6+", where the "+" comes from the many corrections,
    > additions, deletions, etc. that appeared on the mailing list after
    > draft 6 was posted and that were necessary to make draft 6 workable.
    > In particular, most are using the new opcodes
    > because the opcodes published in draft 6 were a surprise, were widely
    > disliked, and were replaced (twice) in subsequent mailings.
    > There is no approved document that defines version 1.
    > The best hope for interoperabilty is to produce a stable standard
    > draft that can gain a reasonable consensus and then finalize the
    > process.
    > 
    > Bob
    > 
    > On Wed, 18 Jul 2001, Matt Wakeley wrote:
    > 
    > > Robert,
    > > 
    > > Your idea kills any hope of interoperability if some vendors choose to
    > ship
    > > products based on certain revisions of the draft - the current "0 vs 6"
    > case
    > > in point.
    > > 
    > > -Matt
    > > 
    > > "Robert D. Russell" wrote:
    > > > 
    > > > Marjorie:
    > > > 
    > > > True there have been new opcodes, but there have been new opcodes
    > > > before.  My point is why start changing the version number NOW
    > > > when we haven't been doing it before?  By your reasoning, we should
    > > > be up to version 7 now, not version 2.
    > > > A problem with changing the version numbers is that the current
    > > > scheme by which an initiator offers versions to a target is that
    > > > there can be no holes in the offering.  If the version numbers
    > > > change too quickly it will be a lot of work to track the
    > > > intermediate versions.  A version change should be really significant,
    > > > ie. at the IETF level, not at the draft level.  We are still at the
    > > > draft level.
    > > > Bob Russell
    > > > 
    > > > On Wed, 18 Jul 2001, KRUEGER,MARJORIE (HP-Roseville,ex1) wrote:
    > > > 
    > > > > >  My personal opinion is still that draft 7 is really just
    > > > > > a refinement and clarification of ambiguities in draft 6, and does
    > > > > > not add any major features that justify a version change. However,
    > ...
    > > > >
    > > > > Not true, there are significant changes to opcodes and some change
    > to header
    > > > > fields between v6 and v7 - that should *at least* be a criteria for
    > a
    > > > > version number change!
    > > > >
    > > > > Marj
    > > > >
    > > 
    


Home

Last updated: Tue Sep 04 01:04:16 2001
6315 messages in chronological order