Lists Home |
Date Index |
I agree, but a schema doesn't have to travel with
the instance in an export/import scenario.
It just has to be agreed on or otherwise,
the fillintheslots won't be there or won't
be square (names, types, sizes, etc.). It
sounds simpler than it is for reasons most
people on this list are surely familiar with.
Import/Export doesn't need hyperlinks but
it does need a schema in some form for a
bare minimum of contract agreement.
The message content is the two-headed gator.
No one can solve that without apriori agreements.
Everyone who has done distributed computing
knows it, hates it, learns to negotiate and
write a tight spec for it or cite one.
Then pick a transport. SOAP plus the web
security languages appear to be a good way
to have a boilerplate section for where
people now write "assuming a WAN/LAN connection
and a common directory location".
From: Paul Prescod [mailto:firstname.lastname@example.org]
"Bullard, Claude L (Len)" wrote:
> It can't be done ad hoc without standard discoverable
> vocabularies. Even REST can't do that. That's why the
> major attention can't be SOAP or REST; it has to
> be schema development.
SOAP and REST are obviously at two different levels (syntax and
architecture) but nevertheless you MUST work out some basic parts of
your processing model before you start your schema development. I think
that REST works out of the box with basically any schema but I'm not
100% sure on that. I *am* sure that using either component RPC or
one-way messaging paradigms will make it impossible for you to use
hyperlinks and that is a major factor in schema design. If you do not
design your schema specifically for RPC or one-way messaging, the schema
is likely NOT to work properly with them because well-designed schemas
tend to use hyperlinks to reference external information.