Re: NEW ISSUE: Define "ought to"

[Willy Tarreau <[e-mail]>]

Hi Julian,

On Tue, Jul 30, 2013 at 09:57:11PM +0200, Julian Reschke wrote:
> On 2013-07-30 21:43, Mike Bishop wrote:
> >If I'm following you correctly, this could be restated:
> >
> >·?MAY? ? Strictly at your discretion based on what matters to you
> >
> >·?ought to? ? Or your implementation will be less effective/efficient
> >than it could be, without hurting anyone else
> >
> >·?SHOULD? ? Or your implementation will cause peers / the network to
> >suffer for your stupidity
> >
> >·?MUST? ? Or you won?t be able to interoperate with anyone
> 
> Almost. Violating an "ought to" may hurt others, jusr not in the way 
> BCP14 says.

I'd say that "ought to" here in the HTTP spect is generally a good friend's
advice from some other implementors that got trapped and know how to avoid
this. There's nothing normative in what follows "ought to" so those who
won't follow it will not cause harm and might only suffer themselves.
SHOULD is a MUST with an exception if you know you can safely ignore it. 

Anyway the difference is quite clear to me, and really useful.

Willy