(a) SOGS will always be soggy, regardless. It's got such a stinker
of a name even the guy who coined it can't stand it :-).
(b) SOGS doesn't have to solve the naming problem, since SOGS
presumes a world in which it, and similar thorny issues, have been
solved. That's not a cop out - it's a statement that this and similar
problems need to be worked on, will be worked on, are being worked
on. But then what?
In this sense it's quite like what Infospheres is trying to do -
reason about what things look like into the future, without necessarily
having to figure out all the plumbing that gets us there. There
are far too many details for anyone to even get them all in their mind
at the same time, let alone resolved. The trouble with most of us
is we get too caught up in our little pieces of the puzzle to ever
try to contemplate the bigger picture of where it's all going when
you put the pieces together.
Now that's not to say that the issues Gordon & Ernie were/are
grappling with are not relevant - they most certainly are. And
at some point I'm sure we'll touch on them over in SOGSland.
One day I'll probably write some monstrous position paper disguised
as an innocent distobj post on how interface repositories,
traders - as in ANSA or CORBA, a general predicate calculus -as in SQL,
web search engines and indices- your basic AltaVistas and Yahoos,
LDAP servers, DNS, and URLs as arbitrary resource specifiers,
perhaps with an inheritence scheme as you guys suggest in Infospheres,
are really all early precursors to some unified distributed, replicated,
scalable resolution service which superimposes all of these things.
But rather than focusing on how to get it just right, I'll probably
concentrate more on (i) why it's needed, and what form it needs
to have (ii) scenarios of how it gets used.
I reserve my 'getting it just right' fixations only for those parts
where I simply can't stand the mediocrity everyone else seems to
push forward, such as on the basic connectivity protocols.
Without the right comms, we can't even get to square one of
building the first experimental shadows. Whereas a general
solution to naming/resolution, as important as that is in due course,
doesn't hold up initial shadow construction. At least I don't think it does.
>Back to ONESIMUS, Ernie's vision the Sunday of our fateful physics
>library alcove brainstorming session. It is a virtuouso performace in
>acronymization, at the very least. Enjoy...
>
>> From ernest@pundit.cithep.caltech.edu Sun Apr 23 18:15:06 1995
>> To adam@cs.caltech.edu
>> Subject: Beyond ONE
>>
>> Okay, I had another epiphany in the shower.
Neat. I find the best stuff happens in the shower too. Other good
places: driving, sleeping/dreaming - yes I actually *dream* of this
shit - scary huh?, doing some urgent task that
needs both hands, like diapering :-), shovelling snow, or pushing
loaded supermarket carts.
Worst possible place to think of new stuff, or even remember good
stuff you already thought of, and swore you weren't going to forget:
sitting at a keyboard.
Actually, I can usually remember the general
ideas hours and even days later. But I never get the exact wordings-
those are lost forever. And somehow the replacement paras. that come
spilling out later are never as satisfying as those precise combinations
of words that laterally popped into your brain in the shower or in bed
or whatever inconvenient place they showed up.
Does this all sound familiar to everyone?
>We may just have to
>throw out the plumbing (and replace it with *TP), or we may have to
>chuck the operating systems (and replace the with oSpace), or we may
>have to replace the programming models (and replace them with Cells), or
>we may have to rip out the file systems (and replace them with ROOFS),
>or we may have to migrate to a new platform completely (Munchkins,
>anyone?), or some combination therein, to free ourselves from the
>limitations of personal computers networked over end-to-end Internet
>protocols...
Yes, yes, yes, yes, yes, and yes. Is this really news Adam?
Without even addressing the jargon of your suggested replacements,
the fact that a major architectural overhaul is still required, on pretty
much all fronts, shouldn't be a surprise. All the more reason to suggest,
as JoeK has, that it's still too early to simply give in and be resigned to an
http-based web. We can, and must, do it all properly,
right up from about IP, I think.
Until and unless it's all predominantly async and multicast, it's
just never going to scale up to SOGS, or whatever you want to call
the endgame. And until and unless every last bloody object/document/
component/shadow/whatnot, from the protocol objects right up to
the domain objects, doesn't have every aspect of itself all
metadata'd up and queriable and introspectable, the whole mess
of it will never stick together.
Ron.