[omniORB-dev] Re: [omniORB] Implementation Repository

[kendall bailey]

Thomas Lockhart wrote:

> If it is implemented in python (which sounds good to me; note that 
> omniIDL requires it anyway and we haven't yet demonstrated how to use 
> this with another orb) then we should probably use an XML 
> representation for the flat file since the interfaces ship with python.

The fact that omniIDL is implemented with Python shouldn't constrain the 
implementation of the ImR.  If I use Java or C++ and only the standard 
CORBA APIs, then the ImR should be portable to other ORBs which don't 
use Python in their IDL compiler.  I plan to use Python just because 
Python is fun and omniORBpy is a great ORB.

I think the details that Duncan has spelled out as far as the encoding 
of data in the object ids and how the ImR would interact with a server 
process to handle the indirect binding has already shown how this could 
be used with another orb.  Are there still open questions?

I'm not convinced that XML is appropriate for ImR data storage needs. 
 My experience with it has been that parsing it is quite slow.  I like 
what omniNames does for persistence.  It's simple and it works.  I'd 
though I might try to generalize the approach for the ImR persistence 
needs.  In particular I'm not sure how to have a well formed XML 
document that can have data added incrementally just by appending more 
lines of text.

>
> Will be posting comments. The IDL posted by Kendall has a lot of 
> content not directly related to fundamental ImR interactions (e.g. 
> load information etc) and I'd like to split the discussion to focus on 
> the fundamental features first...
>
The load balancing stuff is critical for me, which is why I've included 
it in my first go.  I tried to keep it quite simple in the interfaces 
I've defined and allow for varying levels of implementation.  It seems 
natural to me to put it in the ImR since both the ImR and an active load 
balancing service need to do indirect binding.

I'll look forward to your comments.

Thanks,
Kendall