RE: Integrating XP Into Web Infrastructure

Dave Winer wrote:
> Yes, I'm sure you have done it. I could pack up a Frontier table, and
> base64-encode it and pass it to another Frontier node, and they
> would fully understand what they're talking about, but..
>
> What if I want to interoperate with code written Java, for example?

This is really off the point I was trying to make, but to answer the
question, I built a language independent representation and then wrote
language bindings for it. I don't use Java anywhere, so I never wrote one
for it, but it would have been trivial.

Interoperating at the language level requires common representation and data
semantics, not XML specifically. That doesn't mean XML isn't a good way to
achieve this, of course.

> This was the big gotcha in RMI. You had to be in Java to understand it.
> Nice idea for Sun, but no thanks. Theoretically I suppose we could have
> hacked it, but it took less time and was safer to get SOAP out there, and
> coax Sun into supporting it, so that we could do this openly and reliably.

I understand. The point is, we've all been here before. Before RMI there was
DCE and Sun-RPC, and the arguments are the same. And DCE was even an open
standard. And it goes on and on back in time. Microsoft has a whole suite of
language bindings to DCE RPC (via DCOM), and they didn't need XML to do it.
If they rip out the plumbing and stick in SOAP or XP, that doesn't really
change anything except for people that for whatever reason won't use a tool
that's exchanging RPC PDUs but will use one that exchanges XML messages.

> Net-net: If you want to interop, XML is great, that's what it was designed
> for. If you want to get the same thing with HTTP-POST, use XML,
> and then you would have re-invented XML-RPC. ;->

I'm not trying to argue that XP or its ancestors are a bad idea; but it's
not a new idea to build a network data representation standard and then
write language bindings that enforce application semantics.

As an observer, and a consumer of such things, what I'm looking for (and I'm
speaking only for me) is the compelling argument for why I should move to
XML-based exchange of information that I'd never want to read on the wire in
a million years anyway.

With respect to this thread, I'm simply saying that XML does not have huge
inherent advantages over HTTP POSTs for passing structured data *when the
XML exchange is bound to an HTTP POST*. The key is that last part. If the
protocol is transport independent without affecting the way the payload is
represented, then I can see the advantage. Whether it would ever be of use
to me is another matter, but I see the advantage.

--------
  Scott Cantor
  cantor.2@osu.edu          Don't blame me, I voted for Kodos!
  Office of Info Tech                       -- Homer Simpson
  The Ohio State Univ

Received on Tuesday, 9 January 2001 12:14:06 UTC