[
Lists Home |
Date Index |
Thread Index
]
Pipelining is fine. Someday they will discover MIDI. :-)
Arguments based on "RPC is bad for security; REST is good
for security because the programmers will make naive mistakes"
are "slowest runner" arguments. A marathon run with a bad
knee will be just as painful and longer than a sprint.
UDDI needs GUIDs. That may be a bad thing. I think the
WSIO is just trying to make the code on the loading docks
interoperate as advertised. What they do after that depends
on how other organizations do their jobs. It's a cold world.
len
-----Original Message-----
From: Mike Champion [mailto:mc@xegesis.org]
Sent: Friday, March 01, 2002 9:46 AM
To: xml-dev@lists.xml.org
Subject: Re: RE: [xml-dev] REST & types
3/1/2002 10:07:54 AM, "Bullard, Claude L (Len)" <clbullar@ingr.com> wrote:
>
>A lot of the arguments posed for REST seem to depend on programmers
>being lazy. So once again, we get arguments that
>the architecture should be designed for "the slowest runner".
>That is not how competition works. At the root of
>the web service and WSIO work is the realization
>that cooperative competition (sometimes called,
>"smart love") is what we cope with in the human
>side of this, and that the technology is just plumbing.
Hmmm. It seems to me like the argument is that large
scale architectures should be designed for marathoners
rather than sprinters, not that it should be designed
for the slowest runners. Just as sprinting and
marathoning are very different disciplines, so are
desktop computing and large scale distributed applications.
The WSIO clearly has the agenda of teaching the sprinters
to run marathons. I don't know anything about running, so
maybe that works ... and maybe SOAP-RPC really will scale.
I'm not sure who's being lazy, the Web people who don't want
to learn the nasty details of how to make RPC work over
high-latency, unreliable networks or the desktop people who
want to pretend that the Internet is a giant LAN? Anyway,
I'm encouraged by things like XPipe and the XML Pipelining
spec, and hope for new tools that to bring the dataflow model
down to the desktop (not sure ... maybe the BEA Weblogic
Workshop? ). I think we agree that neither RPC or dataflows
are a one size fits all solution.
|
