netconf and python

[Colin Alston]

On 2007/09/19 10:54 AM Bernd Zeimetz wrote:
>> Define "much". Yes, beasts like Zope use a bucketload of memory. I see
>> a couple of python daemons running on my workstation that use about
>> the same amount of memory as a bash shell though ;-)
> 
> But the pure python interpreter needs about 400% of the memory which is
> used by dhclient. This is not a problem for my workstation, but ...

Are you talking about the interactive shell or a basic Python process? 
  If the complexity of the design will better suit Python then 
reinventing the wheel doesn't save anything except waste time.

>> Certainly that sucks for embedded devices with limited memory. Are
>> they going to be a common target for netconf though? (Serious
>> question)
> 
> I think they should be a target for netconf, too.

Please don't be so hasty to make that a requirement. Making something 
a target for embeded devices has a huge number of design criteria that 
would very well take the whole thing back to the drawing board. Even C 
does not save you from the stringent size requirements which would 
horribly hamper something that could be a hugely useful tool to 
desktop and server users instead.
I can say from first hand experience as well (we develop embeded linux 
based routers) that most of these tasks are already stripped down and 
done by custom software. You don't have storage space for the same 
init scripts, you don't have space for most things out there, your 
configuration is stored in often very proprietary NVRAM layouts which 
gets unrolled into ramdisk by custom scripts.
Python may remove this as a target automaticaly (the standard interp 
is far too large to port to most MIPS/MIPSEL embdedded devices), but 
for the converged devices between the really embdeded and true x86 
platform it is totaly fine and lots already use Python as their main 
target (IronPort springs to mind)