Joseph S. Myers wrote: > I note that many of the questions already added to that page could be seen > as suggesting a greater role for the RMs than there is at present in > setting directions for GCC development. RMs (in their RM capacity) do not > set priorities or make plans for GCC development; although we can > determine dates for development stages and branches based on what features > we want to get in them, we can't cause development to be concentrated in > particular priority areas.
We can't direct development per se, but I personally feel that we've been too laissez faire over the past few cycles. I think we could do more to lead. Even the power to say no is a way to encourage development of other things. I've been hesitant to do that for fear of appearing overly strong, but if there's a mandate for that kind of leadership, then I think we could provide more of it. There is, however, a "business model" question as well. GCC does not have a single driving organization behind it; the silicon companies are interested in optimizing for their CPUs, Google is interested in general functionality and performance on its platforms of interest, and Red Hat and SuSE are focused on GNU/Linux features and performance. Questions like "what's the plan for plugin infrastructure?" implicitly ask "who benefits enough from plugins -- whether in a monetary sense or just because they care enough -- to build plugin infrastructure?" One of my current goals is to understand what things are possible in GCC and to drive funding for those features. I gave a talk at the Linux Foundation Collaboration Summit recently where I specifically highlighted plugin infrastructure and link-time/whole-program optimization as things I saw as potentially very valuable. -- Mark Mitchell CodeSourcery m...@codesourcery.com (650) 331-3385 x713
