On Sep 25, 2008, at 09:17 , Dave Thomas wrote:

> Just what was wrong with Test::Unit? Sure, it was slightly bloated.  
> But it seemed to work.

Saying Test::Unit was slightly bloated is like saying RDoc/ri is an  
easy and fun package to maintain. You gave up RDoc/ri because you  
didn't want to maintain it anymore and Nathaniel did the same.

miniunit

    lib  loc     646 total
    test loc     905 total
    totl loc    1551 total
    flog = 711.105369946191

test/unit

    lib  loc    3571 total
    test loc    2464 total
    totl loc    6035 total
    flog = 3103.27739878118

(numbers not totally current but very close)

Time to run 10,000 passing tests:

test/unit       : 0.782527s
miniunit v1.2.1 : 0.340559s
miniunit v1.3.0 : 0.467056s

Time to run 10,000 failing tests:

test/unit       : 12.205469s
miniunit v1.2.1 :  0.839336s
miniunit v1.3.0 :  1.133835s

(numbers are completely up to date)

> Why not remove mini/xxx for now and package it instead as a gem. Let  
> people play with it for a while, and let it stabilize. Then consider  
> integrating it into a future Ruby release.

This makes me feel like I've been standing in the corner talking to  
myself this whole damn time:

2006-10-30 - miniunit 1.0 released as a gem
2007-11-08 - miniunit 1.1 and proposed it to ruby-core [13528]
2007-11-16 - Akira approved miniunit in core [13586]
2008-06-09 - miniunit 1.2 released
2008-06-11 - I asked for approval _again_ [17200]
2008-06-16 - approved by matz [17275]
2008-06-16 - kouhei pipes up with some good concerns [17276]
2008-06-17 - miniunit 1.2.1 released
2008-07-13 - kouhei withdraws his objections [17753]

I've mailed core. I've mailed talk. I've released gems. I've blogged...

I've asked for feedback every way I know how trying to reduce pain and  
complications for when it came time to integrate into 1.9. Kouhei was  
the only one to step up with serious objections.

I know I sound like a broken record on this topic, but I'm really sick  
of all the bloat that comes along in ruby. For a language well known  
for its elegance and expressiveness, stdlib does a horrid job of  
leading the way as a shining example of this. How many options  
processors do we need? (a: 2. 1 actually got removed from 1.8) How  
many copies of a half busted ruby parser do we need? (2: irb & rdoc -  
neither have been reengineered to use ripper). How many test  
frameworks do we need??? TWO apparently. runit is still in 1.9. ftools  
vs fileutils... etc etc etc etc...

I'm trying my best to actively combat this bloat and get things back  
to the lithe elegance that I fell in love with back in 2000. The sad  
thing is, every single example above was true back in 2000... we've  
gone 8 years and still haven't had a serious overhaul on stdlib. The  
fact that there is no lib/deprecate.rb should be evidence enough. I  
love ruby, but I worry that the weight of the bloat is pulling itself  
down.