> From: Chen, Weijing [mailto:wchen@tri.sbc.com]
> It is indeed more complex than what I described as I simplified the case a
> bit. The moral of story is that the complexity and cost of re-engineering
> pre-standard solution to standard solution must be reasonable enough to
> justify the migration from both vendor and service provider aspects. The
> most of software cost is development and testing. Therefore, simple to
> implement and straightforward to test MUST have higher priority over other
> nice-to-have features such as processing efficiency, bandwidth efficiency,
> etc. Hey, we have gigahertz processor, gigabits memory, and gigabit
> network
> everywhere (sort of).
I don't think that you can make blanket statements that "aspect X of the design" must be higher than "aspect Y of the design" of the design. I think you need to keep all the development goals in mind and design as best you can and recognize where and when the trade offs need to be made.
Certainly for your problem set, simple to implement and test could be a higher priority. That may not be a priority for other implementations. Not everyone has a gigahertz processor, and even if they do most have the processor to deal with the bytes on the wire and not for network management functions. Hey I still want this to manage my toaster (just joking -- but that was a touted benefit of SNMP way back).
One of the biggest reasons that standards are hard is that one person's high priority item is someone else's "I don't care".
Cheers, /gww