Re: [ogpx] VWRAP is going no where.

[Joshua Bell <[e-mail]>]

On Sun, Jan 10, 2010 at 10:18 PM, Morgaine
<[e-mail]>wrote:

> On Wed, Jan 6, 2010 at 4:52 PM, Joshua Bell <[e-mail]> wrote:
>
> I'm aware of a fair amount of *implementation work* happening right now,
>> and Meadhbh is *cranking on revisions to the LLSD/LLIDL draft*. Yes, it's
>> a bit quiet here, but I'm not too disturbed; there's a lot of
>> writing/implementation to do to catch up to the state of discussions.
>>
>
> Joshua, that answer just adds fuel to the worries.  Referring to the
> highlighted sections:
>
> 1) *Implementation* of what?  In order to implement a protocol, one must
> have a protocol to implement.  It is one of the worries in this group that
> instead of working out in the group what the protocol needs to do, one or
> two parties may be attempting to impose a de facto standard to override
> group discussions.
>

Implementing what's in the current drafts, and implementations in different
languages. The implementations used/shared previously targeted older drafts.


Also (as Mike points out later in the thread), prototype implementations are
input into the standards process, and should fuel discussion. They do not
magically generate consensus, although they may certainly influence it by
showing running code. That said, while I wish I was aware of some prototypes
of that nature, I'm not.




2) Cranking out *revisions of LLSD/LLIDL* based on what discussions?  The
> latest detailed discussions about LLSD/LLIDL were in MMOX.  What is being
> revised?
>

Meadhbh received feedback from several implementers in private communication
(including some internal at Linden), and some of the feedback from back in
the MMOX days still needs to be addressed. She's attempting to consolidate
that feedback into another draft ready for further feedback from the group
and prototype implementations.

The specifics I'm aware of are allowing for LLIDL to express unidirectional
messages and updates to the variant syntax. There were also some concerns
raised about loss of precision while round-tripping real values through
text-based serialization, but those were resolved without spec changes.