From: Mark Baker (distobj@acm.org)
Date: Wed Mar 15 2000 - 16:55:32 PST
At 04:33 PM 3/15/00 -0600, Jeff Bone wrote:
>
>Hell, yes. :-) I'd go one further: not only is it not necessary for
every mime
>type / application set to have its own protocol, I firmly have come to
believe
>--- in part because of what I saw going on in the IETF --- that a
*minimalist*
>HTTP, perhaps with just "GET," would be sufficient for anything.
How would you ever change anything? I could see GET/POST or GET/PUT, but
just GET?
>I used to
>argue the "protocol per application domain point-of-view. Argued for
protocols
>that were *like* HTTP but *weren't* HTTP. I've come to believe I was
entirely
>wrong!
Amen brother!
>
>With HTTP, you've got a generic request-response system --- i.e., message
>passing. Message passing is provably sufficient for anything computable.
So are smoke signals. 8-)
>But
>wait, there's more! With "servers" on both ends, you can have callbacks,
so you
>can do asynchronous stuff. With XML and something like XML-RPC, you've got a
>typed RPC mechanism and on-the-wire object format. You want objects? Fine,
>build an interface repository and reflection on top of that.
Except that then you're using HTTP as a transport protocol, not a transfer
protocol. Kinda defeats the point. If you can do everything with HTTP what
does RPC buy you and why does it need to be anywhere, let alone on top?
>The punchline to all of this is, I actually think DAV is rather an abortion.
>Ick. The more crap like that we cram into our star protocol, the less
tenable
>it becomes.
For the most part, I'd agree with you; most HTTP extensions I've seen are
so much crapola because the extension methods make no sense in the
architecture
of the Web; document transfer representing computational state (REST[1],
ala Roy).
But WebDA does. Its methods are completely architecturally aligned with REST
(unlike arbitrary interface-level API RPC cruft, like an application method).
Having said that, I don't see a great need for WebDA because, IMHO, the cost
of deployment is much greater than the benefit; you can do most of the good
stuff with just HTTP 1.1.
>This kind of stuff should live above the wire, indeed, above HTTP. If all
>distributed systems can be implemented in terms of RPC, and HTTP provides a
>*standard* and rather friendly RPC just with GET, and XML provides the
>marshalling format, then why re-invent the wheel?
Whoa, "RPC" has no place in that paragraph. Hopefully you meant
"message passing".
BTW, I've finally sparked some discussion on xml-dist-app@w3.org[2].
You might want to sign up. I've managed to convince one of the core
developers on Casbah (http://www.casbah.org) that he doesn't need RPC
and can do what he wants to do with just HTTP.
[1] http://lists.w3.org/Archives/Public/xml-dist-app/2000Mar/0038.html
[2] http://lists.w3.org/Archives/Public/xml-dist-app/
It's funny how these discussions happen at the same time, but independantly.
WebBroker was submitted to the W3C as a note at almost the same time this
stuff clicked for me in May 98. Weird.
MB
-- Friends don't let friends do RPC
This archive was generated by hypermail 2b29 : Wed Mar 15 2000 - 17:06:42 PST