SG10 had its first face-to-face meeting on April 16. It was attended by:

Clark Nelson (chair)
Jens Maurer (scribe)
Benjamin Kosnik
John Spicer
Tom Plum
Tana Plauger

Jens' detailed notes of the discussion are on the wiki at: http://wiki.edg.com/twiki/bin/view/Wg21bristol/SG10
I plan to post these as a separate message, for those who don't have access to the wiki.

Here I summarize the major decisions reached, addressing the questions I posted in January.


        How closely should our recommendation match clang?

We decided to go for distinct object-like macros, rather than a scheme based on a function-like macro. We like the granularity that clang provides. It was noted that, when a feature is refined, we can increase the value of the object-like macro. We haven't made a decision yet about whether the sequence of values should be serial numbers (1, 2, 3, ...) or based on dates of adoption; we'll make that decision based on a concrete example, when we have one.


        Should there be feature-testing for library features?

We decided that there should be. We're especially hoping for feedback from Benjamin on the granularity. We also decided that it would be valuable to cover at least some C++11 features, especially ones that aren't yet universally implemented.


        What form should our results take?

At this time, we plan to publish our recommendations only as a WG21 paper. The document will include rationale and a description of the criteria we use. To give the recommendations credibility, we should at appropriate times get votes of PL22.16/WG21 endorsement for them.


        Next steps

This we didn't discuss much at the meeting, but I think we really need to have a recommendation ready for a vote at the Chicago meeting in September. I'll be working on something that will cover all the stuff that was just voted in. I hope to put it in the post-meeting mailing, but even if I miss that, I'll have it fairly soon. We'll have to discuss it, especially granularity, and refine it. I'll arrange teleconference(s) as necessary, but hopefully we can do most/all of it on the reflector.

--
Clark Nelson            Vice chair, PL22.16 (ANSI C++ standard committee)
Intel Corporation       Chair, SG10 (WG21 study group for feature-testing)
clark.nelson_at_[hidden]