Re: Objectives discussion

[Andy Bierman <abierman@cisco.com>]

At 12:26 PM 11/21/2001 +0100, Juergen Schoenwaelder wrote:

>>>>>> Andy Bierman writes:
>
>Andy> If we have to choose between common sense engineering and
>Andy> academic purity, the choice should be obvious.
>
>I am not sure what this sentence really tells me - but being paid as
>an academic, I somewhat tend to disagree with it.


fair enough -- I'll take it back.
I guess I'm talking about the design tradeoffs that were
discussed in London.   One could optimize the data definition language
for the benefit of a few MIB compiler writers at the expense of a
20,000 MIB writers and a 1,000,000 MIB readers -- or the
other way around.

Not everyone agrees with all the 'accepted' objectives.
Some objectives are likely to lead to solutions which
conflict with other objectives. The document doesn't even
make it clear if we should optimize for the largest benefit, 
or for the largest installed base, or for the lowest
development costs, etc. The objectives document is no
replacement for a functional spec.

For example, (a new one!) I think it would be a disaster to
build syntax extensibility into the language.  Isn't it bad
enough we have all these vendor MIBs implemented instead of
the standard MIBs? I sure don't want vendors to be able to
create their own SMI, via extensions.
 From an academic POV - a cool feature; 
 From an operations POV - a potential nightmare.


>/js 

Andy