On 3/15/2014 9:20 AM, joel jaeggli wrote:
On 3/14/14, 1:34 PM, Joe Touch wrote:
..
The doc includes a number of SHOULDs that are underspecified, e.g.,
SHOULD implement a timeout on retransmissions (what value?), and SHOULD
abort (under what conditions?)
I sympathize with the authors lack of desire to specify specific values
which may be appropriate now but not into the future. That said. I would
be happier with examples. (I am aware of some of them)
My basic metric on SHOULD is that it needs to explain the reason for not
using MUST or MAY. I.e., there's a recommendation here, but not enough
to be invariant.
If a SHOULD is 'SHOULD do X', that implies there are reasons not to;
that warrants an explanation.
If a SHOULD is "SHOULD be capable of X", that implies both enough of a
target that it warrants a described procedure to get there AND reasons
when this might not be appropriate.
If you don't know enough to answer either of these, then you don't know
enough to say "SHOULD".
Joe
