On 03/10/13 13:31, Dicebot wrote:
My role as review manager does not matter here. Responsibility of review manager is to collect opinions and information from community, he does not have any authority of his own.
It's not about your opinions of the code per se, so much as about your experience of what is likely to provide a smooth review process that is less likely to end in rejection for trivial rather than major reasons (meaning a second round of review and therefore more work for you).
Remembering std.serialization discussion though, it felt like people are more in favor of merging generic functionality into matching modules instead of keeping everything self-contained. You can also just do pull requests for smaller parts without any review if they are decoupled from std.rational - but be warned that waiting until those are reviewed and merged by Phobos devs may take a considerable time ;)
Yes, that's my fear. :-(
In the end it is your own judgment. There are no strict rules and it is always good to see a module with personality behind ;)
I think I will follow Jacob's suggestion to make private all that is not _essential_ to have in public, with the scope to pull these out into broader Phobos later. That should allow the essentials of std.rational to go forward without disturbing anything else, and then its generically useful stuff can be parcelled out later without delaying the module itself.
And of course if there is consensus during the review process that certain things _should_ be made public and moved elsewhere, I'll do that.
