Dave,

I like your enhancements. I vote for moving forward with this new definition.

Allen Luniewski
IBM Cross Brand Services
IBM Silicon Valley Laboratory
555 Bailey Ave.
San Jose, CA 95141

408-463-2255
408-930-1844 (mobile)

"Dave Berry" <daveb@nesc.ac.uk>

"Dave Berry" <daveb@nesc.ac.uk> 10/10/2007 01:47 AM

To Allen Luniewski/Almaden/IBM@IBMUS, <ogsa-d-wg@ggf.org> cc "Treadwell, Jem" <jem.treadwell@hp.com> Subject RE: [OGSA-D-WG] Fw: Definition of data consistency

I like this approach.

Is it worth embellishing it as follows:

Data consistency: An instance of data in one resource in a distributed system is said to be “consistent” with one or more other instances of that data elsewhere in the system if it is up-to-date with respect to those instances.

In a given system a policy may define the rules for determining the extent to which data is up-to-date, and for maintaining that degree of consistency.

The changes are:

paragraph 1 - to talk about different resources in a distributed system (putting some context on "data instances", which is not a term we define).

paragraph 2 - to introduce the notion of a degree of consistency rather than simply consistent/inconsistent.

I don't feel strongly about either of these - what do others think?

Dave.

Please see Jem's note below - he can not post to this DL.

I think that his suggestion is a good one.

Allen Luniewski
IBM Cross Brand Services
IBM Silicon Valley Laboratory
555 Bailey Ave.
San Jose, CA 95141

408-463-2255
408-930-1844 (mobile)

----- Forwarded by Allen Luniewski/Almaden/IBM on 10/09/2007 09:49 AM -----

"Treadwell, Jem" <jem.treadwell@hp.com> 10/09/2007 09:40 AM

To Dave Berry <daveb@nesc.ac.uk>, Allen Luniewski <luniew@almaden.ibm.com> cc Subject FW: [OGSA-D-WG] ***SPAM*** Definition of data consistency

Allen/Dave, I can’t mail the data mailing list, so please feel free to pass this on if you think it’s worthwhile.

I agree with Allen about sounding specific, and this is a hard thing to turn into a direct definition without using some term such as “indication.” You might think about using an indirect form:

Data consistency: An instance of data is said to be “consistent” with one or more other instances of that data if it is up-to-date with respect to those instances.

In a given system a policy may define the rules for determining whether data is up-to-date, and for maintaining consistency.

Jem Treadwell
Software Engineer

HP ESS Software
856.638.6021 office | 856.638.6190 fax | jem.treadwell@hp.com
6000 Irwin Road | Mount Laurel | NJ 08054
http://www.hp.com/go/vse

Dave,

I think that I am okay with that suggested change. I am a little bit worried about using the word "measure" as it can be taken to imply things that we really do not want to address (e.g., a means to measure and a quantifiable scale for measurement). I thought about suggesting changing "A measure" to "An indication" but that seems just a bit too vague to me. I don't have a better suggestion right now.

Allen Luniewski
IBM Cross Brand Services
IBM Silicon Valley Laboratory
555 Bailey Ave.
San Jose, CA 95141

408-463-2255
408-930-1844 (mobile)

"Dave Berry" <daveb@nesc.ac.uk>

"Dave Berry" <daveb@nesc.ac.uk> Sent by: ogsa-d-wg-bounces@ogf.org 10/08/2007 02:11 PM

To <ogsa-d-wg@ogf.org> cc "Treadwell, Jem" <jem.treadwell@hp.com> Subject [OGSA-D-WG] ***SPAM*** Definition of data consistency

Jem queried our definition of data consistency, which is currently:

A policy that specifies or describes how up-to-date an instance of data is in one service of a distributed system, with respect to one or more instances of that data elsewhere in the system.

Jem said (in e-mail):

my thought was that data consistency would be a state of the data - it's consistent w.r.t. some other set of data or it isn't; the policy would be the rules that determine whether it is, and what the requirements are.

I've had a look through our document and I agree. We refer to data consistency policies but also to data consistency requirements, mechanisms and strategies.

So, how about this as a revised definntion:

"A measure of how up-to-date an instance of data is in one service of a distributed system, with respect to one or more instances of that data elsewhere in the system."

Best wishes,

Dave Berry
Deputy Director, Research & E-infrastructure Development
National e-Science Centre
15 South College Street, Edinburgh, EH8 9AA
+44 131 651 4039 www.nesc.ac.uk
--
ogsa-d-wg mailing list
ogsa-d-wg@ogf.org
http://www.ogf.org/mailman/listinfo/ogsa-d-wg