>>"My general observation is that a SHOULD in a document
>>is underspecified unless there's a condition that indicates
>>when the action asked for is not required."

That hits the nail on the head.

>>a SHOULD as a MUST when I am on the sending side,
>>and a SHOULD as a MAY when I am on the receiving side.

This is good and very considerate.
Unfortunately, it is the use of an unnecessary SHOULD which
creates the necessity for your considerate behaviour.

And being considerate takes more effort - in software development
time and in CPU cycles.  I wish it wasn't necessary.

Regards,
Attila



-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Marc 
Petit-Huguenin
Sent: 18 July 2008 17:36
To: Iñaki Baz Castillo
Cc: [email protected]
Subject: Re: [Sip-implementors] SHOULD not MUST (was RE: Setting anonymous From)

Iñaki Baz Castillo wrote:
> El Friday 18 July 2008 17:38:44 Attila Sipos escribió:
>> Thanks Michael,
>>
>> This is useful to know.  Many people have misunderstood SHOULD 
>> including myself !!
>>
>>>> 3. SHOULD   This word, or the adjective "RECOMMENDED", mean that there
>>>>   may exist valid reasons in particular circumstances to ignore a
>>>>   particular item, but the full implications must be understood and
>>>>   carefully weighed before choosing a different course.
>> I'm really not sneering or anything but maybe the SHOULD description 
>> MUST include a sentence saying:
>>
>>     "there may exist valid reasons in particular circumstances to use a
>>      SHOULD instead of MUST, but the full implications must be understood
>>      and carefully weighed before doing so.
> 
> I have an easier solution:
> 
> If RFC's would include just MUST instead of SHOULD and MUST then 
> implementation would be easier, interoperatibility better and life happier.
> 
> Sometimes it seems that a RFC uses SHOULD instead of MUST because the 
> author is not totally sure about what he's writting.
> 

I think the best advice for authors was from Henning Schulzinne:

"My general observation is that a SHOULD in a document is underspecified unless 
there's a condition that indicates when the action asked for is not required."

I also remember reading or hearing somebody say that all SHOULD should in fact 
be conditional MUST, which I agree with.

The problem is that a lot of implementers takes SHOULD as a way to write sloppy 
code.  I even read about vendors that refuses to fix a bug because the spec 
says SHOULD and not MUST.

As an implementer, I always try to combine the SHOULDs with Postel's 
Prescription ("Be liberal in what you accept, and conservative in what you 
send.") by considering a SHOULD as a MUST when I am on the sending side, and a 
SHOULD as a MAY when I am on the receiving side.

-- 
Marc Petit-Huguenin           [                                 ]
Home: [EMAIL PROTECTED] [RFC1855-compliant space for rent ]
Work: [EMAIL PROTECTED]            [                                 ]
[                                                               ]
_______________________________________________
Sip-implementors mailing list
[email protected]
https://lists.cs.columbia.edu/cucslists/listinfo/sip-implementors

_______________________________________________
Sip-implementors mailing list
[email protected]
https://lists.cs.columbia.edu/cucslists/listinfo/sip-implementors

Reply via email to