COM/CORBA/DCOP (was: Hello people. I have some questions)

[Duncan Grisby]

In article <9mvkvr01db at enews2.newsguy.com>,
 Alex Martelli <aleax at aleax.it> wrote:

>> I don't know why people would consider CORBA too heavyweight for
>> connecting desktop apps -- the few performance comparisons between COM
>> and CORBA I have seen show CORBA to be as fast or faster than COM.
>
>For *in-process* interaction?!  There's something askew here... I
>did the measurements myself back around 1994/1995, under Windows/NT,
>using COM (in-process and out-of-process) and Iona's Orbix, which
>Iona claimed was the fastest ORB for NT, and the difference in
>terms of performance overhead was *at least* an order of magnitude
>in favour of COM (using C to program "toy"-level servers and
>clients, instrumented for measuring communication overhead, and
>striving to simulate the kinds of loads our own applications would
>place on a componentization infrastructure).  Have things changed
>so drastically since then?

Well, Orbix has always been one of the slowest ORBs, so I don't know
how they can have claimed it was the fastest. I suppose that as long
ago as 1995 there wasn't much competition.

Regardless, the performance comparisons I've seen were talking about
out-of-process use, but not necessarily across a network.

A good ORB will turn an in-process call into little more than a
virtual function call, so the overhead is minimal there, although
probably a bit more than COM. Given that CORBA, largely in-process, is
fast enough for a whole window system (Berlin), it's certainly fast
enough for the sorts of things required by embedding.

Cheers,

Duncan.

-- 
 -- Duncan Grisby  \  Research Engineer  --
  -- AT&T Laboratories Cambridge          --
   -- http://www.uk.research.att.com/~dpg1 --