[lug] can't make this stuff up, folks...

[Bear Giles]

On Mon, Oct 19, 2009 at 6:01 PM, Zan Lynx <zlynx at acm.org> wrote:

> For example: what good does it do to prove that some program to parse
> email addresses is correct if the programmer and proofer did not realize
> that user+tag-tag at example.com <user%2Btag-tag at example.com> is a valid
> email address?


That's a problem with the specification.  All the proof can ever do is prove
that the specification was properly implemented and anyone trying to use it
for more is mistaken.

It sounds flip but it really isn't.  In fact I think it's one of the biggest
problems I (and many others) face.  We're so used to having poorly defined
requirements that we have to trust our own judgements on what's really
meant... and sometimes we're wrong.  Sometimes we're right for the wrong
reason.  (Prime example: naive leap-year code that was correct for 2000.)
Either way we catch the blame when problems occur.

In the ideal world we could demand better specifications.

In the real world...
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lug.boulder.co.us/pipermail/lug/attachments/20091019/927f8aa0/attachment.html>