Jim, thanks for the comments, there are always different ways to look at
things which is what makes life interesting!

On Wed, 2008-12-31 at 07:34 -0800, Lux, James P wrote:
> 
> 
> On 12/31/08 1:38 AM, "Bob Cowdery" <b...@g3ukb.co.uk> wrote:
> 
> > Rob, Frank
> >
> > Thanks for the input. SOAP, usually over HTTP, is as you say very
> > lethargic and entirely unsuited to this type of system. It's also a
> > pretty low level of abstraction.
> SOAP can be at whatever level of abstraction you want. HTTP is just the
> transport mechanism, and I've seen some very fast implementations within a
> single processor (after all, how much does it take to do a "GET").  Where it
> gets clunky is when there's a mismatch between server and client, or where
> there's a lot of "late binding" (i.e. Lots of back and forth to establish
> the vocabulary of the conversation, then once both sides have agreed on what
> they're going to send, actually sending the data.)  A lot of implementations
> use a very "interpreter-like" front end to do the lexical analysis and
> parsing from a not particularly efficient data structure.  Sort of like
> saying "Here's a BASIC program to tell the server what the client needs" and
> then having the server parse the BASIC program.
> 
I've used it to an extent in commercial system where I wanted to
orchestrate across enterprise boundaries using UDDI as a location
service - for that its good but I don't like it much for distributing
the components of a single application.
> 
> 
> >CORBA I believe is pretty much dead
> > except for a bunch of niche users.
> 
> One of those niches, is, oddly enough, software defined radios.  OMGs
> Software Communications Architecture (SCA) uses CORBA for middleware (google
> CORBA SCA for lots of stuff).  So does the DoD Joint Tactical Radio System
> (JTRS)..
> 
> Both of these are fairly big efforts (multi billion dollar), and one can
> find much to gripe about with both, but, as far as installed base, they're
> pretty dominant in the SDR world.
> 
Yes I agree they are major users, in fact most users are now in the
technical arena. I think once you buy into the technology in a big way
its hard to back out or even switch vendor as interoperability was never
that great between vendors.
> 
> 
> There are some good alternative
> > commercial offerings but I'm looking for open source as all my stuff is
> > open source. I have a test bed in pylink-sr and as Ice has Python
> > bindings this would be a low-cost way to get a comparison. At the moment
> > I am using net-jack for audio distribution and Pyro for control
> > distribution. Moving both of those over to Ice would give me a good
> > side-by-side comparison of performance and relative code complexity. It
> > will be interesting if nothing else.
> >
> > 73
> > Bob
> > G3UKB
> >
> > On Tue, 2008-12-30 at 16:07 +0100, Frank Goenninger wrote:
> >> Hi Bob, hi Rob,
> >>
> >> Web Services based on SOAP (or HTML, which means the same, normally)
> >> is all over the place these days. If you require real performance I'd
> >> strongly recommend to stay away from anything XML or CORBA. XML is way
> >> to inefficient in its gross/net information ratio. CORBA requires
> >> significant amounts of processing power to handle all the proxy and
> >> interface finding stuff.
> >>
> 
> Kind of depends on where you need the performance.. If it's something like
> tuning a radio every second, no big deal.  If it's running a FIR filter,
> probably not.
> 
One of the things I want is a proper event service and that's hard to do
with HTTP. I'm prepared to give Ice a shot at being a better CORBA, if
it doesn't get there I've just lost a little time, but either way I get
to add something to the armoury or write something off.
> 


_______________________________________________
FlexRadio Systems Mailing List
FlexRadio@flex-radio.biz
http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz
Archives: http://www.mail-archive.com/flexradio%40flex-radio.biz/
Knowledge Base: http://kc.flex-radio.com/  Homepage: http://www.flex-radio.com/

Reply via email to