90 days est mort. Vive le 90 days.
Arthur A. Gleckler
(29 Jan 2021 22:24 UTC)
|
Re: 90 days est mort. Vive le 90 days.
John Cowan
(31 Jan 2021 00:31 UTC)
|
Re: 90 days est mort. Vive le 90 days.
Arthur A. Gleckler
(31 Jan 2021 01:05 UTC)
|
Re: 90 days est mort. Vive le 90 days.
Marc Nieper-Wißkirchen
(31 Jan 2021 09:21 UTC)
|
Re: 90 days est mort. Vive le 90 days. Lassi Kortela (31 Jan 2021 17:58 UTC)
|
Re: 90 days est mort. Vive le 90 days. Lassi Kortela 31 Jan 2021 17:58 UTC
John: > In my case, at least, it's not so much that the SRFIs are > unpolished, as that they only handle part of the problem, the > part I myself can see clearly. The SRFI discussion provides for > the transition from mere opinions to actual positions that can > then be debated and hopefully resolved. That is indeed the main value of the SRFI process, but if the scope of a SRFI is too broad or the problem is too vast to be understood in a reasonable amount of time, we don't have the time and energy to do a proper review. The scope can be hard to get right. Empirically, reducing the scope has consistently led to more successful SRFIs. Arthur: > Yes, I agree, and I apologize for the unpolished language of my > message. IMHO your message was fine. John: > (Of course there is also the problem that people don't notice > there is a SRFI they want to comment on until they hear that it > is about to finalize. I don't know what to do about that. > Likewise there is the problem of insufficient resources.) Arthur: > I keep thinking that there must be a way to help solve some of our > problems with bribery. More stickers, anyone? Unfortunately bribery cannot solve exhaustion. Our regulars worked very hard all 2019 and 2020; by the second half of 2020, it seems most of us had become overwhelmed. My unpopular suggestion is to stick close to the core language in the relatively heavyweight, big-design-up-front RnRS and SRFI processes, and use more lightweight approaches for ordinary library design. Marc: > The old language in the SRFI process seemed to imply that, for quality > reasons, the 90-day deadline is needed. I'm not at all convinced about > this, so I like your new language better. I see that some kind of limit > is useful because the more SRFIs in draft status, the bigger will the > administrative overhead be. > > In any case, I do not see time constraints as a major issue with the > current SRFI process. The problem I see is that many SRFIs are > discussed by far too few people. Moreover, the people that have been > involved in discussing SRFIs only reflect a small part of the Scheme > community. The R6RS process was blamed for this, but the same can > probably be said about the current SRFI process. I haven't made any > statistics but I have the strong feeling that the community was *much* > more diverse during the first 50 or 100 SRFIs. Good points. It comes back to the problem of not enough people/energy. > In a discussion about the inclusion of SRFI 88/89 to Chez, Kent Dybvig > once wrote: "[...] SRFIs are not standards and are vetted, for the most > part, only by people who are interested in the mechanism." I think this > hits a nail on its head. Our reviewers are all biased. Who does review > and examine thoroughly a SRFI they are not really interested in? If someone is truly apathetic about some part of the language, maybe it doesn't matter if they don't review a SRFI about that part. However, if someone is has a strong _negative_ interest, their comments would be very valuable to hear if presented constructively.