Re: Object-Archiv

[Don Yacktman]

On Thursday, July 3, 2003, at 03:17  AM, Stefan Böhringer wrote:
> I'm in favor of such a software collection.

Me too, and that should come as no surprise.  ;-)

> Perhaps I can put into this discussion some experiences gained from the
> famous MiscKit that was in place in the NeXTstep/OPENSTEP days. You can
> still obtain information about that project on http://misckit.com 
> [Don, if
> you are on this list, please chip in].

I am on this list, but have been on vacation and I just got back.  The 
timing of this thread to coincide with my being away is quite ironic...

> One conclusion is that definitely cvs access is a must. Don Yacktman 
> did an
> excellent job over many years to apply patches in his source tree and 
> bring
> out releases frequently [this was in times when anonymous cvs was 
> alien].
> Then NeXTStep/OPENSTEP ran into trouble and Don didn't have the 
> capacities
> to do the laborious maintenance job any more which brought the project 
> to a
> halt.

Yup.  Too much for one guy, especially once I stopped being a student 
and had to actually start working for my food... obviously CVS is part 
of the answer to that, these days.  The other thing that got in the way 
was writing Cocoa Programming.  That took a couple years...and left me 
with zero free time.  But I felt that the book (and the info it would 
contain) would be more valuable to the nascent Cocoa community than the 
MiscKit at the time (you have to move beyond beginner status before 
you're ready for a big kit that adds to the already huge Cocoa).  Now 
that the community is more grown up, it's time for the MiscKit 
again...overdue for it, really.

> The MiscKit was used commercially, but there were restraints on 
> account of
> two things: the arcane MiscKit license and quality. LGPL is fine but 
> quality
> would still be an issue for a similar GNUstep related project. Such a
> library would potentially grow fast and uncontrolled bringing with it
> duplication of functionality, an inhomogeneous class hierarchy and the 
> like.
> A solution discussed for the MiscKit was to split the library into two 
> part:
> the wild and the tame (the attributes were not used back then). "Wild" 
> can
> be extended in a very liberal way and is considered "low quality". If 
> a set
> classes is to migrate to "tame" it should fulfil several criteria. E.g.
> there have to be regression tests for the classes, they have to fit 
> smoothly
> in the class hierarchy as present, there must be documentation and so
> forth...
> The bottom line is that migration would have to be discussed 
> explicitly and
> decision should be on several shoulders (one is not enough).

That's a pretty good summary of where I plan to head with the MiscKit 
when I revive it.  There's a lot more than just the above though, and a 
few details will be slightly different.  Given what's been discussed in 
this thread, I think everyone here will be happy with what I have in 
mind, if I (and the team I'm assembling) manage to implement all the 
features I envision.  (The final vision is much more than just a 
library, that's for sure.)

Since having me as a bottleneck CLEARLY won't work, I am quietly 
recruiting a board of people who would all have CVS commit ability, and 
work together to keep things in order, and make the releases.  Then if 
any one person gets busy, the others can take up the slack and things 
move on.  I certainly don't feel any need to exercise dictatorial 
control over the kit, by any means.  Back in the day, it worked well 
enough, but it just doesn't scale.  I figure a model pretty similar to 
how Apache is developed is a good way to go there.  (A few things would 
be a little different for practical reasons, but that's pretty much the 
model I like.)

Anyway, my company (illumineX) is helping out with the hosting and 
infrastructure, but we're sort of swamped at the moment.  That means I 
don't quite have a time frame yet for when you'll see kit 3, or the new 
infrastructure open to the public, but I really hope it is before 2003 
finishes out.  (And by all rights it certainly SHOULD happen much 
sooner than that.)  We were going to try to have sometime for WWDC, but 
it didn't happen, unfortunately.  (I got really sick about a month 
before WWDC and that pretty much shot our plans for that completely.)

Once everything is set up, though, things will progress pretty quickly. 
 I feel it is important to take the time to carefully lay a good 
foundation that will be able to go on a lot longer than the foundations 
of the original MiscKit could handle, rather than rush something out.  
(ie, learn from past mistakes and make it orders of magnitude better).  
So setting up the infrastructure isn't going to happen overnight.  If I 
could have my way, though, you'd see something before Panther releases.

Oh, and since this group probably cares about this...initially the kit 
will might only build on Mac OS X, since that's what I use every day.  
I have full intentions of supporting GNUSTEP and would hope to have it 
building and working there ASAP, probably with a bit of help from the 
community, of course.

Right now I don't have a good machine to put GNUSTEP on, so I'll need 
help from the community for that support.  If I could run GNUSTEP, then 
I'd have GNUSTEP support in there right out of the gate.  I do consider 
it vital for the long term success of the project.  If we get a GNUSTEP 
person on the team from the start, support out of the gate might be 
possible.  Any nominations?  <grin>

The thing is, it has to be someone who has demonstrated some talent and 
is willing to put in some time and really wants to be part of the 
project.  Since I don't want to have a lot of people with commit access 
at the start (only 5 or so at first), it is important that those who 
have commit ability actually use it...until things stabilize a bit.  
Then we can look at adding more people with commit ability later.

I know that the above is sort of sketchy, but at least you know sort of 
what I'm thinking (there's a TON more than what I wrote above, of 
course; a lot of details).  I'd like to give time frames, but right now 
my economic situation is pretty tight.  That means I don't have as much 
free time as I'd like to have, and makes it hard to predict when I'll 
have the initial (very, very modest) release out the door.  Best I can 
say is "as soon as I possibly can"...

And as to the library vs. index thing, there are good arguments for 
doing it each way, and it makes sense to me to have both.  I have plans 
to address both; MiscKit first, for the library, and then of.org for 
the other...

I think that covers what's been said in the thread; most of the issues 
that have been raised are things I've already thought through at great 
length and tossed around with various friends to find workable, good 
solutions.  I'm happy to answer questions you may have (if I can), but 
there's a lot of details I can't talk about just yet, while I'm still 
getting them ironed out.  But not that I'm not on vacation, my latency 
in responding should be a little better.  ;-)

--
Later,

Don Yacktman
don@misckit.com