Re: Status of recirculation ballot
On 2015-04-14 17:22:37 +0200, Vincent Lefevre wrote:
>>> There seems to be a problem with the first paragraph or §8.3:
>>>
>>> A flavor may provide the decoration com with the following
>>> propagation rule for library arithmetic operations. In an
>>> implementation with more than one flavor, each flavor shall
>>> do so.
>
> After some thoughts, this doesn't seem to make sense. The
> specification of a flavor doesn't depend on an implementation.
That sentence is indeed an issue, and we debated it extensively in
the context of Motion 42. I've also found Motion 61 which makes
"com" required in all flavors -- but I see no final disposition of
this motion! (In fact, 59-62 are all unfinished as far as I can tell.)
My interpretation of the discussion about all-flavor literals this year
was that this was about an editorial reorganisation, and not about a
substantive technical change. But we seem to have lost track of what
the commonly-agreed-to intent is for the status of "com".
Here is a quote from Christial Keil's 16 Feb 2014 YES vote for M0061:
|
| 1) The revised text lists the com decoration to be *required* for a
| flavor in several places, while the decoration clause (8) says a
| flavor *may* provide it and only *requires* it from all flavors
| of an implementation if the implementation provides more than
| one flavor. Either the current text or the decoration clause
| (8) should be changed to be consistent.
Apparently we never acted on this!
On 3 April 2013, John Pryce wrote:
| - Making "com" optional for single-flavor implementations seemed
| more trouble than it's worth, so it is mandatory in all cases.
| The current rewrite makes §10.11 "Notes on the com decoration"
| look out of place however.
During the M0042 discussion (Jan 2013) we did discuss various implications
of optional "com", especially in the context of newDec().
I also came across a discussion of the possibility that "com" might not be
the strongest decoration in certain flavors, which would affect compressed
intervals (but not in the set-based flavor).
IF we let "com" be optional, it is conceivable to have a flavor that does
not support it (as opposed to an implementation), and such a flavor would
simply be forced to be exclusive; it could never participate in a multi-
flavor implementation.
So perhaps we should go back to Vincent's original correction, and fix the
Clause 8 sentence that makes "com" optional in a single-flavor setting.
Since that paragraph simply contradicts earlier statements we may get away
with claiming this to be editorial cleanup.
Michel.
---Sent: 2015-04-14 16:41:37 UTC