Cameron Laird's personal notes on COM
COM is an acronym for "Component Object Model". Expect considerably
more content in this space in the second half of 1999.
Home
Page for Microsoft COM: COM, DCOM, COM+ Technologies
Don Box is tops. Everything he writes about COM is trustworthy.
[Review books.]
I'm planning articles on DCOM for Unix (and elsewhere)
in 1999 and 2000.
Watch this space for pointers.
Much of what I know about COM is disappointing. David Shepherd launched
a quite literate
thread
which aptly summarized real-life experience with COM. At
a lower level, the annoyances I experience with COM include
multiple ideas-that-might-have-sounded-good-but-sure-aren't-working-out:
- uuID (mysteriously interacting with the Registry)
- object-orientation that's just a burden
- confused equations of class-code and instance-data
- the frequency of such "closed" (mega-)controls as
IE windows with scrollbars which refuse to be
addressed programmatically;
- the whole interface advertisement model, which,
at best, assumes waaaay too much about the information
content of a type specification (to quote Alexandre
Ferrieux, "what MS tool should know the
usable range of a preemphasis factor for
an FFT, just based on its 'float' type???". Alexandre
summarizes: "the self-documentation mantra of COM is
a gross lie. What you can get from the 'comment'
field of an interface/method/class like
void Foobar(...)
is most often ... 'Fooes the Bar'").
- humanly-unreadable marshalled data
The obvious technical competitor to COM (or perhaps DCOM)
is CORBA. On the other hand, in
late 1999 it appears that Microsoft might decide
to abandon DCOM entirely in favor of SOAP [explain].
Cameron
Laird's personal notes on
COM/claird@phaseit.net