Re: [tcpm] [Editorial Errata Reported] RFC7323 (6798)

[t petch <[e-mail]>]

On 27/12/2021 17:51, [e-mail] wrote:
> They are interoperability issues as follows:
>
>> The Timestamps option may appear in any data or <ACK> segment, adding
>
> This implies that receivers MUST NOT treat segments with missing or present timestamps as an error.

I struggle to see why this does not come under the general presumption 
of being liberal in what you accept.  If anything, it is a MUST or 
SHOULD for the receiver and no MAY is needed.

Tom Petch





>
>> The timestamp clock may be derived from a system clock that is
>
> This implies that receivers MUST NOT treat timestamps that look like the current time as an error, e.g., they MUST NOT assume the timestamp starts at a random offset.
>
>> A random offset may be added to the timestamp clock on a per-
>
> This implies the converse of the previous, i.e., that the receiver MUST NOT treat timestamps that don’t appear “close” to their current value as an error.
>
> This is why RFCs MUST (IMO) indicate a rationale whenever keywords are used - especially for wiggle-words (MAY, SHOULD and their converses), noting particularly where exceptions are known appropriate.
>
> Joe
>
> —
> Joe Touch, temporal epistemologist
> www.strayalpha.com
>
>> On Dec 26, 2021, at 11:34 PM, Yaakov Stein <[e-mail]> wrote:
>>
>> None of the three cases I called out will cause interop problems, but they open potential security or privacy issues.
>>
>> But the lack of interop concerns did not dictate capitalization in other cases, such as
>>    A TSecr value received in a segment MAY be used to update
>>    the averaged RTT measurement
>> which is a purely local matter.
>>
>> Y(J)S
>>
>> On Mon, Dec 27, 2021 at 8:08 AM Carsten Bormann <[e-mail] <mailto:[e-mail]>> wrote:
>>> On 2021-12-27, at 06:16, Yaakov Stein <[e-mail] <mailto:[e-mail]>> wrote:
>>>
>>> In any case, the same RFC uses the normative "MAY" in several other places,
>>> so these are probably mistakes.
>>>
>>> For example,
>>>    The three-byte Window Scale option MAY be sent in a <SYN> segment by a TCP.
>>> and
>>>    This option MAY be sent in an initial <SYN> segment.
>>>
>>> So why
>>>    A random offset may be added to the timestamp clock on a per-connection basis.
>>> ?
>>> Is this may to be considered weaker than the other two?
>>
>> Is this an interoperability “MAY”, i.e., does the peer need to be prepared for this behavior?  The other two clearly are, while this seems like local matter.
>>
>> Grüße, Carsten
>>
>>
>> _______________________________________________
>> tcpm mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/tcpm
>
>