Re: [rtcweb] I-D Action: draft-ietf-rtcweb-audio-02.txt

Coming from an implementor's perspective (typically), and having read quite
a few RFCs, I strongly concur with your statement, about using SHOULD in an
RFC requiring spelling out the circumstances where it's reasonable to
ignore.

As Martin said, and I agree, usually implementors take liberties to ignore
the "SHOULDs, RECOMMENDEDs, MAYs, and OPTIONALs" until they get bit by
something (usually in interop).  Then they curse the spec for using non
mandatory language for whatever bit them :-)

Matthew Fredrickson

On Fri, Sep 6, 2013 at 5:27 PM, Harald Alvestrand <harald@alvestrand.no>wrote:

> On 09/06/2013 11:42 PM, Mo Zanaty (mzanaty) wrote:
>
>> True. The same can be said about every SHOULD, but we still bother to
>> specify it.
>> Maybe we should update RFC 2119 to state:
>>
>> Obey: "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT"
>> Ignore: "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", "OPTIONAL"
>>
>
> At the time 2119 was written (I remember discussing this with Scott
> Bradner several times), the intent of the text for "SHOULD" was "you'd
> better do this if you don't have a real good reason why it's not a
> reasonable thing to do in your particular case".
>
> Quoth:
>
> 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.
>
> Of course, people who don't want to cover all the SHOULDs have pushed
> towards interpreting it as "weak recommendation that we can ignore if we
> feel like it", but that wasn't the original intent of 2119, and some of us
> still want to have that word available for use in the stronger meaning.
>
> At one time there was push towards saying that "if you write SHOULD in an
> RFC, you need to spell out the circumstances where it'll be reasonable to
> ignore it - otherwise, write MUST or MAY". That push has petered out at
> this time, but I felt sympathetic to the idea.
>
> That's why I'm so reluctant to use the word in cases where I think a large
> part of the implementors are going to ignore it, and where (in my opinion)
> no great harm comes to interoperability when they do.
>
> But this instance not something I storm barricades over; if the consensus
> of the WG is to use "RECOMMENDED" rather than "recommended", I'll note that
> I'm the rough part of the consensus, and live with it.
>
>
>
>
>> -----Original Message-----
>> From: Martin Thomson [mailto:martin.thomson@gmail.**com<martin.thomson@gmail.com>
>> ]
>> Sent: Friday, September 06, 2013 4:38 PM
>> To: Mo Zanaty (mzanaty)
>> Cc: Jean-Marc Valin; Bo Burman; rtcweb@ietf.org
>> Subject: Re: [rtcweb] I-D Action: draft-ietf-rtcweb-audio-02.txt
>>
>> On 6 September 2013 13:28, Mo Zanaty (mzanaty) <mzanaty@cisco.com> wrote:
>>
>>> If the will is hostile to other codecs, this text won't matter.
>>>
>> The natural disposition of the implementer is hostility toward more
>> work, so I don't see the text making much difference either way.
>> ______________________________**_________________
>> rtcweb mailing list
>> rtcweb@ietf.org
>> https://www.ietf.org/mailman/**listinfo/rtcweb<https://www.ietf.org/mailman/listinfo/rtcweb>
>>
>
> ______________________________**_________________
> rtcweb mailing list
> rtcweb@ietf.org
> https://www.ietf.org/mailman/**listinfo/rtcweb<https://www.ietf.org/mailman/listinfo/rtcweb>
>