Re: [vwrap] Status and future of the VWRAP working group

[peter host <[e-mail]>]

Agreed too.

I'm not involved in the VWRAP working group but I've followed it with
interest. I'll give more of an "external" and "candid" opinion.

First, there is nothing wrong in re-purposing work (#2, #3). We might
once have thought that LL could be a key player in VW research, and
obviously they aren't anymore as most the motivated/interesting
Lindens have left the LL boat anyway. VWRAP is very much marked by
(now legacy with Humble, most likely) LL views, and they ditched the
baby.

Second, I happen to have implemented XML as soon as it appeared in
1998 for the INSEE (the main French Statistics Governmental
organization) in order to exchange data between services, and we had
tried SGML before. We eventually managed it, and designed a pipeline
based on that standard. But very soon we found ourselves trapped in
layers of complexity (xsl, xpath, namespace hell...), big bosses
liking nothing more than presenting *their* bosses a nice reassuring
report assessing that all they do is 100% standard based... Whether
the standard is appropriate not being an issue, and at the cost of a
huge developer/maintainer/education overload. As it came out, XML was
fine for a very little subset of what we did, and horrible for most of
the rest (especially tabular data and http transactions).
Then JSON came out in 2002/2003. At first, our hierarchy thought it
was a joke. But after 1 year of developing a parallel pipeline, JSON
based, they were with us. Same with Ecmascript. We ditched java
although most of the French ministries were engulfing in it (and still
suffer from it). By the way, Ecma initially is a 1guy, 4weeks project,
and Crockford did JSON in a few days. How much effort took XML ? XML
at least taught us one thing in my team, it's that it was not needed
in most of our use cases (which *is* important to know), though we
*did* manage a lot of document oriented data.

Reason why I do agree with Christa. What we need most is a SSO
standard. Corporations need to have their networks of trust, and to be
able connect them together. This is exactly, on another level, what
Hypergrid is addressing. And It's lightweight to the core, so it's
good standard material.

No decent, open economy can develop without SSO. What we're seeing now
is the proliferation of mostly closed appstores/contentstores... of
all sorts. Because these models bring accountability and enable
enforcement. (at the *ghastly* cost of monopolies). Time for a
standard. In the same line of thoughts, securing web services is the
main focus of the 'Harmony' version of  TC39 - ECMAScript.

Second area (yes i know, it's not VWRAP related) that will likely have
be addressed is re-purposing LSL to something more inter-operable,
which can run in lightweight, platform (ie mono) independent VMs too.
llHTTPrequest is fine, but well... We've all tortured it already.
LSL's event-driven, finite state machine is brilliant. Strong Type,...
not so brilliant. Interop, awfull. Most flaws relate in one way or
another to the built-in requirements of a money module. Which brings
us back to SSO.


Enough rambling on my part.
Last, I wish to thank all the people who've really worked hard on
this, because it's always very tough when key initiators and founders
of a technical project withdraw for non-technical, but
political/commercial reasons.

Reading you all is always a pleasure.



On Sat, Jan 15, 2011 at 3:14 PM, Morgaine
<[e-mail]> wrote:
> That's well said.

>  All we ever needed was a means of
> associating an agent ID with a protected credential,

yup


> On Sat, Jan 15, 2011 at 1:23 AM, Cristina Videira Lopes <[e-mail]>
> wrote:
>>
>> I'm leaning towards #2 and #3 simultaneously :)
>> Let me explain.

yup