Re: Apple's Darwin commitment

Am 14.09.2005 um 04:13 schrieb Kevin Van Vechten:

> On Sep 13, 2005, at 6:46 PM, Markus Hitter wrote:
>
>> In the Mac OS X 10.0/10.1 days, Darwin was, updated in CVS  
>> regularly. One could watch source code being developed in small  
>> changes, compile the stuff weekly, etc ...
>
> By the time I had come to Apple, CVS was already offline, but I  
> made it a priority to get source snapshots out promptly with every  
> release, organized by release.

Exactly, and clearly less often than before. Not your fault, though.  
You did the best the situation would allow you.

>> With 10.2 and 10.3, public source were updated for official Mac OS  
>> X releases only. If something stopped working between releases,  
>> you had to wade trough a huuuge patch. But still, a standalone OS  
>> could be compiled successfully, resulting in more than one  
>> OpenDarwin release.
>
> Are you speaking from experience?

Yes. With 10.0 and 10.1, I managed to get a kernel running on my PPC  
604 box, because I could patch the kernel in small steps from the  
already working Public Beta one.

The single big patch from 10.1.5 to 10.2 securely prohibited this  
strategy. It took more than a year for some people far better than me  
to understand what's going on in this kernel and to finally make a  
patch. Again, it turned out to be a very minor change, possibly a  
typo in the kernel source.


>> With 10.4, the do-it-yourself OS is obviously gone.
>
> I totally disagree.  Every source tarball and patch that I used to  
> build Darwin 8.0.1 has been published.

So why, by all means, does it take more than a week for a knowing  
person to do the exactly same again? Rob stated more than once the  
8.0.1 release is unstable and not re-buildable. What's the difference  
between the stuff on your platter compared to what Rob has?

> Shantonu and I have been working on DarwinBuild which is a much  
> more robust build environment for Darwin than was ever available  
> for Darwin 6.x or 7.x.

Yes, that's great. But how does it fit the picture "Don't expect to  
find a working OS, don't expect single Darwin projects to work on Mac  
OS X"?

> For the past couple weeks there have been several volunteers  
> completing near-world builds of Darwin almost nightly.  If you're  
> interested in the recent progress, check out <http:// 
> release.opendarwin.org/Leimy-Logs4/>.

Again, this is very good news.

If somebody like Joshua is so desparately seeking for a working  
Darwin OS as a foundation for his research, what stops you from  
telling him about such efforts? Instead, he's listening to something  
like "You're welcome to have a look, but don't expect anything  
working". "Disencouraging" is the only word I can find for such answers.

> As far as I know, this is the biggest coordinated darwin building  
> activity that has ever happened.

Maybe there's going on a lot in some Blogs or on IRC or wherever.

On the other hand, I'm sure there are many more people like me who  
mostly listen to what's going on. Occasionally they pop in, want some  
task getting done, and go back to listen mode a few days or weeks later.

Other, similar sized projects, seem to get quite a benefit from these  
people. They are very helpful to expose bugs, spread the word,  
testing different environments. Often, they provide a test case or  
even a patch and, surely you agree, discussing a patch is much easier  
for the maintainers than to find, debug and fix the issue themselves.

If you want Darwin to catch these people as well, and you want to  
expand Darwin over these few highly encouraged all-time contributors,  
you really should find some way to inform the interested audience  
about current activities.

I tried to do my part by reading Darwin and a few other mailing lists  
daily, give interested  people hints in the (hopefully) right  
direction, tracking OpenDarwin's home page and occasionally trying  
some piece of Darwin software. This isn't much, but better than  
nothing(?).


Now I find myself too far off line.



Am 14.09.2005 um 05:42 schrieb Jordan K. Hubbard:

> It's always easy to look back at the past and say that everything  
> was much better, cheaper, more well-constructed, etc. and  
> conveniently edit the bad parts out of your memory, and I suspect  
> that more than a little of that is occurring here.

I consciously avoided any word about the quality of this older stuff.  
Past is past and the only times talking about the past is useful is  
when you either want to avoid making mistakes twice or you want to  
extrapolate into the future.

> I was there, and I don't recall ANYONE "compiling stuff weekly",

To be precise, I waded through CVS kernel sources using date tags.  
Recompiled and tested the thing for each single line of change, if  
possible. This narrowed and finally fixed a bug and the useful  
results were exposed to the public.

This was not exactly computer science, but it did the job.

> I mostly remember them complaining that they couldn't get anything  
> to compile at all. :-)

Like today? Just kidding.


> There's a darwin release in progress as we speak, in fact, all  
> driven by opendarwin folks.

How comes you missed this important fact in your previous post?

> There are no black helicopters buzzing around here, Markus.  Just a  
> number of engineering realities and the sorts of compromises that  
> need to be made when dealing with the melding of two or more  
> dissimilar cultures. If you or Joshua want a monoculture, [...]

I'd really like to see OpenDarwin going beyond the stage of tracking  
Apple's changes.


Markus

- - - - - - - - - - - - - - - - - - -
Dipl. Ing. Markus Hitter
http://www.jump-ing.de/