Re: [mmox] Creating walled gardens considered harmful

[Morgaine <[e-mail]>]

On Sun, Mar 29, 2009 at 11:45 PM, Jon Watte <[e-mail]> wrote:

My feedback is that we should make the implementation as cheap as possible
> to get some end-user visible benefit for virtual world interoperability, so
> that we can get lots of worlds on board. Agreed?
>
>
Not agreed at all actually. :-)  Although we haven't got around to talking
about the parameters of our solution space yet, I think it's always a good
time to reflect on the IETF's mission
statement<http://www.ietf.org/rfc/rfc3935.txt>,
because this sets our solution parameters as well:


In attempting to resolve the question of the IETF's scope, perhaps the
fairest balance is struck by this formulation: "*protocols and practices for
which secure and scalable implementations are expected to have wide
deployment and interoperation on the Internet, or to form part of the
infrastructure of the Internet*."


Our work here is completely for nothing if it doesn't scale for wide
deployment on the Internet.  Anyone can knock up a simple bit of interop for
an easy perceived win, but success would then be an illusion that shatters
as soon as the number of users grows.

It can be hard to impress upon people the extreme difficulties that arise
when services are scaled into the millions of users and beyond without an
adequate design in place, but once you've experienced it once then you know
you don't ever want to experience it again.   A technically cheap design is
very often a non-scalable one which puts you on a "*You can't get there from
here*" dead-end road when you start hitting its scalability limits.  If
there is a single thing that we must do right in MMOX, it's to not propose
cheap but non-scalable solutions.

Linden Lab has a registered user base of over 16 million users and well over
80 thousand users online concurrently, so it won't surprise anyone that
scalability became an issue a long time ago.  Indeed, the AWG was founded
with scalability as a primary requirement on day 1, and when LL+IBM brought
this effort to the IETF, it was with very large numbers in mind for the
interoperating metaverse of worlds.

In this context, proposing a technically cheap solution that has not been
assessed for scalability has somewhat limited interest.  Interop has the
natural and expected consequence of growth in numerous dimensions (eg. user
population, concurrent users online, concurrent communication channels,
number of worlds, people within a single region, number of objects, etc),
and our solutions must reflect that.  The design of OGP was predicated on
it.

This may mean that some easy solutions are given the thumbs down in our
context, but that's not really a problem:  you don't need a *Massively
Multi-participant* IETF standard if you just want a cheap, low-scalability
solution.

Of course, if you can find a cheap solution that also happens to scale well,
all the better!  But usually, TANSTAAFL applies.


Morgaine.








On Sun, Mar 29, 2009 at 11:45 PM, Jon Watte <[e-mail]> wrote:

> Morgaine wrote:
>
>> There are many parts to the puzzle, but participants need implement only
>> those parts of relevance to them.
>>
>
> I agree. However, that side-steps the problem a little bit. My feedback is
> that we should make the implementation as cheap as possible to get some
> end-user visible benefit for virtual world interoperability, so that we can
> get lots of worlds on board. Agreed?
>
> Sincerely,
>
> jw
>
>