Python for large projects?

[Jody Winston]

>>>>> "Janko" == Janko Hauser <jhauser at ifm.uni-kiel.de> writes:

    Janko> Jody Winston <jody at ldgo.columbia.edu> writes:
    >> I've used Python for three large distributed applications (each
    >> > 100k lines of Python).  The only other recommendation that I
    >> would add is to define interfaces to the code.  If this is done
    >> with a product like ILU or FNORB, you can then easily call the
    >> Python code from other languages.  In addition, you can replace
    >> critical sections with other languages if needed.
    >>
    Janko> I have thought about this as the perfect way to do ``data
    Janko> hiding'', publish only special interfaces for your
    Janko> data. Has this a significant overhead, if used on one
    Janko> machiene, or is it overkill just for this purpose :-)?

For my data, which are large arrays read from disk, the ILU program is
50% slower on a simple benchmark that just transfers the arrays from
one address space to another on the same machine.  The xml-rpc
program, which transfers the same data from a client to a server, is 2
times slower as the ILU program[1].

However both of these bechmarks, for my case, do not really mean
anything since the real code is compute bound.

To understand the impact of a distributed object system will have on
your design, you'll need to understand how you will access your data
and the possible bottlenecks.  

Footnotes: 
[1]  That's really amazing in my book since all of the xml-rpc data
is ASCII.

-- 
Jody Winston
Lamont-Doherty Earth Observatory
RT 9W, Palisades, NY 10964
jody at ldeo.columbia.edu, 914 365 8526, Fax 914 359 1631 

Under US Code Title 47, Sec.227(b)(1)(C), Sec.227(a)(2)(B) This email
address may not be added to any commercial mail list with out my
permission.  Violation of my privacy with advertising or SPAM will
result in a suit for a MINIMUM of $500 damages/incident, $1500 for
repeats.