Very well put, Vaughn.

Regarding "service level interoperability", it's not really a subset of VW
interoperability at all, but lies orthogonal to it because it is a property
of all multipart systems that implement services.  I'll try to explain.

Client-server systems for example have at least two distinct parts with
distinct roles, server(s) and client(s).  Service-level interoperability
typically consists of designing and specifying the protocols or interfaces
between them in such a way that any part of the system is interchangeable
with another equivalent part that performs the same role, say from a
different manufacturer, and the system as a whole still continues to work
normally.

This rarely needs to be honored with a fancy title such as "service level
interoperability" in these days of IETF standards and highly cross-platform
web applications.  Historically however, applications did not behave quite
so nicely, and vendors often sought to lock customers in with hidden
proprietary interfaces, so there was a need for adding "service level
interoperability" to the tendering documentation in days gone by, to avoid
surprises later on.

Today, the need for that has almost disappeared, and if your online service
has a documented protocol interface then "service level interoperability" is
virtually assured without doing anything at all, assuming normal commonsense
has prevailed during development (and there is no deliberate obfuscation of
course).  There is only one other area of this topic where a little more
needs to be said.

Protocol messages can normally be validated syntactically to a strong
degree, and whether a message is correct or not is normally known
immediately upon syntactic validation.  Unfortunately, that is not always
the end of the story, because protocols can transport complex data items
which are valid syntactically yet invalid semantically.

This is the last vestige of where that term is commonly encountered,
as *Service
Level Semantic Interoperability*.  Is it relevant to us?  Yes, a little.
For example, it would do us no good to transport Collada meshes from an
asset service to a client that tries to render them as some other graphics
format --- that would create a semantic mishap, because one party thinks
that the items means one thing and another party applies a different
semantic.  So yes, there is a little more for us to consider here, but it's
not a lot.  In most part we have already stated the solution every time that
we have mentioned MIME types for describing content.  This is mostly a
solved problem.

There may be one or two other areas where we have to specify the required
semantic alongside the simple matter of protocol syntax, but that's a
standard part of defining protocols.  There is nothing really new here.

Lastly, service level interoperability focuses entirely on single services
and their clients, so it's unrelated to interoperability between multiple
services, such as a set of virtual world services.  This means that, apart
from defining type semantics, it doesn't get us even a step closer towards
interop between virtual worlds.


Morgaine.






==========================

On Wed, Mar 30, 2011 at 10:07 PM, Vaughn Deluca <vaughn.deluca@gmail.com>wrote:

> Katherine wrote:
> >It seems to me that accomodating "full" interop only would reduce the
> >number of possible implementers and use cases for our work
>
> I am sure that nobody  suggested to we restrict ourselves to "full"
> interop only.
> "Service level interop" is clearly subset of VW interoperability. You
> can't have VW interoperability without service level interoperability!
> However, If i understand Morgaine right, she is worried that the VWRAP
> specs will be *restricted* to service level interop only.
>
> It has been argued (sorry, I forgot by whom) that proper
> specifications of service level interop will give us virtual world
> interop for free. I think we need a bit more, but i really have a hard
> time envisioning how service level interop can be specified in such a
> way that it *prevents* VW interop, at least not if we pay attention to
> this aspect, and clearly we do. So in conclusion i do not see much of
> a problem.
>
> Izzy wrote in another tread "This whole argument between service level
> interop and 'full' interop eludes me." At first it eluded me to, but
> now i think that what is actually expressed in these discussions is
> the question of the scope of our effort as well as design approach.
> Some prefer to work bottom up, following their primary interests in
> getting the low level protocols working, especially since that will
> already be good enough for (all?) of their use cases . Some prefer a
> more top-down approach, first sketching the high level picture of the
> users perception of VW interop,  and from there working downwards,
> obviously also because that approach optimises the realisation of
> *their* use cases.
>
> I think we need both, and above all, i do feel that the two approaches
> are not al all incompatible and both are without any doubt square
> within the current charter.
> Formally splitting our effort in two working parties along the current
> visible fissures and getting each to work on their own interest is a
> recipe for disaster. It will only strengthen the animosity that is
> already slowing us down tremendously, and will sustain the infighting.
> Rather than spitting efforts off, we need to address the causes for
> the current disagreement and found common ground.  In my view that is
> not all all hard.
>
> I have been reviewing the discussion we had in September and i am
> actually amazed at the level of consensus that is expressed in those
> email exchanges. Unfortuanately we have been very bad at consolidating
> that consensus. As a result we recycle the same problems time and time
> again. The archives make very depressing reading from that point of
> view. We need to do more documentation for sure.
>
> I am currently going over the September discussion and looking up the
> places were we all agreed on the way forward.  Much of that is still
> undocumented, and my aim is to get those points written down in the
> wiki.
>
> As i final point i want to make clear how I understand the term
> "Service Level Interop". I used the term, and since the glosary entry
> is still emtpy, i feel obliged to add at least my personal reading. If
> somebody disagrees, please add an improved version.
>
> Service level interoperability:
>        A subset of "Virtual World Interoperability" as defined by the VWRAP
> protocol. Service level interoperabity loosely describes specific
> interactions between VWRAP endpoints. These inteactions form the glue
> that hold a particular virtual world together, but might just as well
> be used for communication between different VWRAP compatible virtual
> worlds (i.e. crossing trust domains).
>
> I intend to put this up in the glossary, but first will see how it
> flies here  :)
>
> On 3/29/11, Katherine Mancuso <kmancuso@gmail.com> wrote:
> > Hi everyone,
> >
> > I want to speak up for agreeing with Meadhbh & Mike about keeping a
> > role for service level interop in this group.  We've made good
> > progress towards this goal and can continue to work on it.
> >
> > Here is an alternative proposal to any which has been brought up so
> > far.  I'm aware this may not be fully correct in its technical
> > details, as I am not a software architect:
> >
> > Could the partisans of "full" VW interop consider working together on
> > a draft specification that is something like the intro or David's
> > piece in scope that lays out which specific capacities would need to
> > be supported at a minimum to allow for "full" interop, perhaps getting
> > input from implementers such as the Hypergrid folks and the folks who
> > coordinated the teleport test at Linden?  Citing existing service
> > specifications (either ones developed within this WG, or outside
> > specifications like XML, Collada, etc) is preferred, and for
> > capabilities for which there are not existing service specifications
> > or the existing specifications don't work well enough, address that to
> > lay out a clear roadmap of what must be developed.
> >
> > My vision here is that folks who are doing service-level work could
> > continue developing and implementing their individual services, and
> > the folks who wish to do "full" interop could define a menu of
> > services which must be implemented for "full" interop.  Implementers
> > could then choose their path based on their use cases: implement the
> > "full" interop specification including all the specifications called
> > for, or implement individual services.  I believe that both of these
> > concepts can exist under our existing charter or with limited
> > amendment to the charter and intro, which is much easier for everyone
> > to agree to than entirely rewriting both, and does not require
> > trashing years of work.
> >
> > It seems to me that accomodating "full" interop only would reduce the
> > number of possible implementers and use cases for our work
> > dramatically, not to mention cut out a productive body of folks in
> > this group who have been contributing an awful lot of documents and
> > implementing.
> >
> > For example, to illustrate my point:
> >
> > From my work as a SL developer focusing in education, I know there's a
> > substantial use case in K-12 education in the US for walled gardens,
> > because schools have big legal liability problems with letting minors
> > wander unwalled virtual worlds (most school libraries in the US have
> > internet filters for the same reasons) and have fairly intense
> > supervision requirements which require substantial moderation &
> > surveillance tools.  However, a shared asset server that contains a
> > core of "safe" content might be of interest to these folks, since
> > educators don't necessarily need to reinvent the wheel for their
> > classrooms every year.  This is a really good case for service level
> > interop ... implement the asset server specification only, not the
> > "full" one that allows you to teleport to other worlds.
> >
> > On the other hand, universities have far greater interest in letting
> > students and professors teleport among university spaces (which
> > happens metaphorically in the real world all the time), and fewer
> > liability issues.  Sharing an asset server might be of interest to
> > them, but so too might "full" interop.  They'd want to implement the
> > "full" interop specification.
> >
> > (Paging Fleep Tuque, are you here to make this case better for me?)
> >
> > TL;DR - Proposal is to amend the charter & intro so that they allow
> > the "full" interop people to work in one community on their ideas and
> > the service level interop people to work in parallel on their ideas,
> > rather than assuming that one model has to exclusively dominate the
> > group.
> >
> > I will be unavailable to post anything else much lengthy through 3 April,
> > FYI.
> >
> > Katherine
> >
> > --
> >
> ---------------------------------------------------------------------------------------------
> > Katherine Mancuso
> >
> > ISIS Inc, Community Manager (http://www.isis-inc.org)
> > Sex::Tech Conference, Social Media Chair (http://www.sextech.org)
> > The Vesuvius Group: metaverse community builders
> > (http://www.thevesuviusgroup.com)
> > GimpGirl Community Liaison (http://www.gimpgirl.com)
> >
> > http://twitter.com/musingvirtual
> > http://facebook.com/kmancuso
> > http://www.linkedin.com/in/kathymancuso
> > Second Life: Muse Carmona
> >
> ----------------------------------------------------------------------------------------------
> > _______________________________________________
> > vwrap mailing list
> > vwrap@ietf.org
> > https://www.ietf.org/mailman/listinfo/vwrap
> >
> _______________________________________________
> vwrap mailing list
> vwrap@ietf.org
> https://www.ietf.org/mailman/listinfo/vwrap
>