ietf
[Top] [All Lists]

Re: Running Code

2009-03-04 11:59:54
Hallam-Baker, Phillip wrote:


I don't see the value of running code quite as others do.

For me the value of running code is that the requirement to actually
implement stuff does tend to reduce the scope for complexity, you have
someone in the room pushing against something that will make work for
them. And the other advantage is that there tends to be a closer
relationship to actual real world needs.

But you do not have to do A to get B and doing A does not guarantee that
you get B.

Another alternative is to require people to produce a proof of
correctness for their protocol. That provides even greater encouragement
to be concise and to get it right the first time.


The running code strategy can also backfire. I have seen groups where
one party has a large development team on call that allows them to drive
the specification. And I have also seen groups where no progress can be
made because the programmer who wrote the dufus code won't allow the
dufus to be deleted from the spec. 

Yes, there is text in the document saying that having early
implementations should not prevent to break compatibility between
two versions of an Internet-Drafts

Coding too early can also be a problem.

This is the main problem.  Nearly everybody agree that early
implementations are good for the development of a protocol.  It's a
fact that early implementations are frustrating for developers so
very few do it.

The proposed solution (mandatory acknowledgement in I-D and RFC) has
been rejected on every comment made in this mailing-list, so this
will not happen.  So what is the solution to encourage developers to
do early implementations?

-- 
Marc Petit-Huguenin
Home: marc(_at_)petit-huguenin(_dot_)org
Work: petithug(_at_)acm(_dot_)org
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf

<Prev in Thread] Current Thread [Next in Thread>