RE: [boost] Re: FORMAL review process not working well?

From: Paul A Bristow (pbristow_at_[hidden])
Date: 2004-10-22 12:01:06

| -----Original Message-----
| From: boost-bounces_at_[hidden] 
| [mailto:boost-bounces_at_[hidden]] On Behalf Of David Abrahams
| Sent: 21 October 2004 19:40
| To: boost_at_[hidden]
| Subject: [boost] Re: FORMAL review of "Output Formatters" - results
| 
| > We got a lot of feedback relating the "Output Formatters" library [1].
| > "Yes" votes : 4
| > "No" votes:   2
| > "Abstain" votes: 2

| > Thus, I wil consider this library "Pending Acceptation". In other
| > words, it's considered Accepted by default, but since it will be
| > redesigned, a new short review will take place.
| 
| I have no opinion whatsoever about the library's merits (I didn't
| look), but what you're describing here surprises me.  The group's
| reception to the library seems from the vote to have been lukewarm at
| best and the library is going to be redesigned.  Of course review
| managers have the perogative to render any verdict they like, and I
| don't know what "Accepted by default" is supposed to mean exactly, but
| it doesn't seem like an appropriate result given what was written
| above.

I am concerned that this shows again that the review process is not working
well.
We are not getting enough reviewers and we are getting Boosters input far
too late.

The number of reviewers is often, as here,
far too small for the votes to be statistically significant,
and IMO a negative impression is often caused by a few vociferous objectors,
whose views I doubt are representative.

We need far more input from the 'lurkers'.

As a group we are only really getting to grips with submissions at the
review stage.
This is the stage when we should only be concerned with a few final details,
but in too many cases major re-designs are proposed,
even when previously there had been general
agreement about the desirability and overall structure.

I see a chicken and egg problem here. Stuff is not subjected to the real
review
- widespread user acceptance - until it is in the Boost library.
But much useful stuff isn't getting that far.
(The poor accessibility of both the 'files area' and 'CVS area' doesn't help
-
I note that much stuff is now being made visible on assorted private web
sites).

It is most unreasonable to expect submitters to perfect both code,
and especially documentation, unless they are confident of ultimate
acceptance.

In practice, many submissions are not really refined until after acceptance.
For example, Jeff Garland's invaluable date-time system.  Much of the
refinement
has come from user use-in-anger experience, not from the review process
itself:
that only established that it was a good starting point. 

Most reviews recommending acceptance end with provisos.

Can suggest again that we split the process into two stages:

1  Acceptance in principle - as suggested by the review leader in this case.

2  Final acceptance for the next release.

Paul

Paul A Bristow
Prizet Farmhouse, Kendal, Cumbria UK LA8 8AB
[phone]  [phone]
mailto: pbristow_at_[hidden]