Iterators/vectors (was: [Biocorba-l] BSANE and bioCORBA)

[Brad Chapman]

[Mike proposes generic any-based iterators]

Alan:
> OK - I'm not a big fan of non-type safe iterators and any's, for a number
> of (possibly biased & incorrect) reasons:
[...]
>  - Using any's suddenly exposes the fact you're using CORBA with its
>    associated helper classes. (I like my CORBA to be invisible).

Martin:
>    It depends. If your implementation deals directly with methods from an
> IDL you must see CORBA anyway, w or w/o anys. But you can make your
> implementation modular - and let only one module deal with the CORBA world
> (together with anys) - and the rest is completely CORBA-unaware.

I tend to agree with Alan on this one. While I do see the point of
more generic iterators (hell, I don't like writing all of those
specific iterator classes!), I think that adding any's makes the code
much more difficult to deal with, especially for those new to
CORBA. If you add anys, then suddenly you have to understand
typecodes and other CORBA internals, as Alan mentions. I'd prefer to
make the IDL as easy as possible to get into, to help encourage new
people to make use of the standard. This is what I like about the
current biocorba IDL -- I was able to start off with very little
initial knowledge of CORBA and able to get going quickly.

>  - I'm not quite sure of the status of 'any' implementations outside of
>    Java. Can anyone comment?

I haven't used anys except in very few cases, but my experience with
some free ORBs (orbit-python and Fnorb) is that it is messy enough
just dealing with "regular" CORBA stuff. I still have Fnorb bugs in
the current biopython-corba implementation that I can't figure
out. I can't imagine it would get any less ugly when we get to more
complicated features :-). 

Brad