From: Matt Armstrong Date: 2002-01-15T02:40:59+09:00 Subject: Re: RCRCR matz@ruby-lang.org (Yukihiro Matsumoto) writes: > Hi, > > In message "RCRCR" > on 02/01/13, David Alan Black > writes: > > |Yes, a Ruby Change Request Change Request. > > My only wish for RCR is that I want to discuss about RCR > > * publicly > * on mail list > > The current RCR is pull model. I have to check RCR periodically > (and often forget). It's little bit annoying and I'd like to > see more public comments. I too like the idea of RCR discussion being mainly list based. Here's a nice "life of an RCR" that I think would work well: 1. Person has idea. 2. Person posts to ruby-talk asking about pros and cons of the ideas, mentioning possible RCR. 3. The idea changes, maybe dies, maybe lives on, after discussion. 4. If the idea lives, person creates an RCR on ruby garden. 5. Initially, RCR lives in "raw" state -- no comments allowed, etc. 6. RCR moderator makes sure RCR lives up to some quality standards (i.e. idea has been discussed on list, RCR clearly describes the change to Ruby and describes the pros and cons, contains links to relevant ruby-talk posts, Matz has not recently said he would reject such a change on the list, etc.) 7. RCR moderator moves RCR to "open" state, and mail gets sent to ruby-talk. 8. Final discussion happens on rubygarden.org, but this will likely be short since all major issues are discussed on ruby-talk already. 9. RCR is accepted or rejected. Another idea is that bi-weekly mailings to ruby-talk could describe the current "open" RCRs. This way, it is unlikely that any RCR will stay open for too long. There are too many undecided RCRs on rubygarden. I basically followed this path when I wondered if Regexp#match should take a "character offset to start searching" argument, as an optimization. It turns out that String#index can do that, so the RCR is unnecessary. However, the discussion led me to discover that a "character offset" is not what I'd want anyway, for performance reasons with encoded character sets like SJIS and UTF8. Another good example is RCR #20 (http://www.rubygarden.org/article.php?sid=69). Another example is RCR #59 (http://www.rubygarden.org/article.php?sid=160). This should have been discussed on the list first. I think it is a good idea, but being posted first to rubygarden.org limits its audience. Examples of recent bad RCRs are #58 and #57. The changes requested are not clearly described, and would have been much better to post to ruby-talk soliciting comments first. So they would have benefited from the above process. -- matt