pausable explanation for the Document Series
Bound, Jim
Jim.Bound at hp.com
Mon Jun 9 16:07:47 CEST 2003
I understand. But keeping bugs out is something I consider pretty easy
to do if we follow a good engineering process and review. Where it gets
tricky is the classic engineering trade-off of risk vs going forward and
where we live with the warts. Another is if we have option 1, 2, and 3
and the WG is split on supporting all options. This gets back to my
issue of we need to put the assumptions we have on the table better.
/jim
> -----Original Message-----
> From: Keith Moore [mailto:moore at cs.utk.edu]
> Sent: Monday, June 09, 2003 12:13 PM
> To: Bound, Jim
> Cc: moore at cs.utk.edu; jari.arkko at piuha.net;
> hinden at IPRG.nokia.com; problem-statement at alvestrand.no
> Subject: Re: pausable explanation for the Document Series
>
>
> > You miss the point I think? And this is exactly how most products
> > ship it is impossible to find all bugs in any first
> software release.
> > What you try to find are the show stoppers based on what that means
> > for that product function.
> >
> > So I suggest your disagreeing with reality?
>
> Not at all. Protocols aren't quite like products in several
> ways. If a vendor makes a buggy product, it screws those
> that buy that product, but not everyone else who uses the
> same protocol. (unless the vendor has a monopoly, in which
> case I'd argue that the vendor has a higher
> responsibility.) Also, protocol bugs are often harder to fix
> than product bugs(depending on the product). Also, protocols
> aren't specified to the same level of detail that the code to
> implement those protocols are, so it's reasonable to expect
> fewer bugs in protocol specs.
>
> Again, I think the attempt to characterize the desire to have
> higher quality in our protocol specs before we call them any
> kind of standard (and before we consider them ready for
> deployment) as wanting those
> protocols to be "bug free" or "perfect" is misleading.
> Clearly it is neither possible nor realistic to expect bug
> free protocols. But that does not excuse our publishing a
> protocol standard with known bugs, nor does it excuse our
> failures to do due diligence to avoid major problems in our
> protocol designs.
>
More information about the Problem-statement
mailing list