From: Jeff Garland (jeff_at_[hidden])
Date: 2004-10-04 15:12:56
On Mon, 4 Oct 2004 22:25:47 +0300, Peter Dimov wrote
> The most important part of a formal review is the accept/reject
> vote. It is often accompanied by comments that (a) act as a
> rationale and (b) suggest changes that may cause the vote to change
> from 'reject' to 'accept'.
Yes, but most reviewers tend to be much less disciplined about their reviews...
> As a convenience, when the issues are minor, the library is not re-
> reviewed. Instead, the review manager declares that the library is
> accepted subject to issues X, Y, and Z. The implication being that
> these issues, if not addressed, would cause the rejection of the library.
Well this is apparently where our understanding differs a bit. I think the
review manager has to be able to ask for required changes to the library that
are independent of whether the issue would rise to the level of 'causing the
library to be rejected'. He also needs to be able to NOT ask for a change
even though a reviewer has requested one -- essentially determining the
relative importance and implications of the issues raised.
Case in point, with Iostreams -- I've asked Jonathon to remove the 3rd party
headers from the library because of the licensing complications it creates as
well as the potential conflicts with user installed libraries. Only one
reviewer commented on this (the conflict issue, not the licensing part -- that
was my concern), but it was my judgement that this is an important change
because of the implications. I believe I would have abrogated my review
manager duties if this issue was postponed till an audit just prior to the
1.33 release. Would I reject the library if we somehow had to have the 3rd
party header -- no, because the licensing is compatible with Boost, but we
would have to have a long discussion on the list about the implications of it
no doubt ;-)
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk