"Georg Wrede" <georg.wrede@nospam.org> wrote in message news:422A08F7.4020807@nospam.org...
> Matthew wrote:
>> Been reading the book - the article itself is available here http://www.catb.org/~esr/writings/cathedral-bazaar/cathedral-bazaar/ - and thinking about how this pertains to D.
>>
>> I am wondering whether we might be able to make the cathedral - the D language - smaller, and the bazaar much larger. It'd need Walter to become more like Linux Torvalds and less like, erm, Walter.
>>
>> I'd be interested to hear the thoughts of anyone who might be motivated to read the whole article as to whether they think this might be applicable to D. (Imagine releasing several def's of the language in a single day!)
>
> I've been thinking for some time on a practical issue, only somewhat related to this.
>
> Currently we have the DMD compiler, and that's it.
>
> What we'd need, IMHO, is another track, with a version of the compiler that is only meant for testing the candidate ideas. To make this absolutely clear, the "toy" compiler would have to have a very distinct name. (Think about proto, test, toying, or "tasting a feature", etc.)
>
> Technically, it would be DMD, no biggie. Slapping on, and off, new features would be somewhat easier on Walter, since the whole idea is to stay pre-pre-alpha with those features.
>
> Actually, if slapping on an interesting feature would require too much work (technically) to cooperate quick-and-dirty with some official feature, then that would just be disabled for that build.
>
> Then we'd see posts by Walter:
>
> """
> Hi, I just made a regexps-in-language version, but disabled wchar and dchar in it. Check it out, is this feature so
> cool we'd consider doing it for real?
>
> Oh, and last week I made another, that supports dscript (not to be confused with DMDscript). But it's broken, I disabled all compiler switches, but it still chokes on shebangs, just so you know.
>
> The former is called Burn-0-115-1-regex.zip and the latter Burn-0-115-1-dscript.zip. There is just the compiler, and regex.zip has a modified Phobos in the package. As normal, slap them in dmd/bin of 0.115 release. That's all to get up and trying them.
>
> Next week I'm trying the 2+ dimensional dynamic arrays, but to save time, I'll implement only 1 and 2 dimensions. If the response is good then I'll try to first figure the general solution, and then, maybe next month do a Burn version. """
>
> The Burn :-) releases would have a non-silencable start-up text, in the spirit of:
>
> This is a feature test bed. DO NOT use this for ANYTHING. Does not come with ANY documentation. THIS VERSION HAS BUGS,
> is incomplete, and we
> know it. Do not use. For a FREE, QUALTY compiler, see
>
>                  www.digitalmars.com
>
>      *** burn-0-115-1-regex.exe, 200503150830, build 4812 ***
>
> ---------
>
> So the Burn series DBURN???.EXE progs would be totally throwaway. And the Linux dburn??? binaries.
>
> ############
>
> Why? What's the point?
>
> Today, Walter has to really think things through beforehand. And do them PROPERLY in the compiler. All this slows down progress.

Nice idea. My concern is that there'd not be time for Walter to do it. If this could be bazaar'd, maybe that'd succeed?

Matthew wrote:
> "Georg Wrede" <georg.wrede@nospam.org> wrote in message news:422A08F7.4020807@nospam.org...
> 
>>Matthew wrote:
...
>>Actually, if slapping on an interesting feature would require too much work (technically) to cooperate quick-and-dirty with some official feature, then that would just be disabled for that build.
>>
>>Then we'd see posts by Walter:
>>
>>"""
>>Hi, I just made a regexps-in-language version, but disabled wchar and dchar in it. Check it out, is this feature so cool we'd consider doing it for real?
>>
>>Oh, and last week I made another, that supports dscript (not to be confused with DMDscript). But it's broken, I disabled all compiler switches, but it still chokes on shebangs, just so you know.
>>
>>The former is called Burn-0-115-1-regex.zip and the latter Burn-0-115-1-dscript.zip. There is just the compiler, and regex.zip has a modified Phobos in the package. As normal, slap them in dmd/bin of 0.115 release. That's all to get up and trying them.
>>
>>Next week I'm trying the 2+ dimensional dynamic arrays, but to save time, I'll implement only 1 and 2 dimensions. If the response is good then I'll try to first figure the general solution, and then, maybe next month do a Burn version.
>>"""
>>
>>The Burn :-) releases would have a non-silencable start-up text, in the spirit of:
>>
>>This is a feature test bed. DO NOT use this for ANYTHING. Does not come with ANY documentation. THIS VERSION HAS BUGS, is incomplete, and we
>>know it. Do not use. For a FREE, QUALTY compiler, see
>>
>>                 www.digitalmars.com
>>
>>     *** burn-0-115-1-regex.exe, 200503150830, build 4812 ***
>>
>>---------
>>
>>So the Burn series DBURN???.EXE progs would be totally throwaway. And the Linux dburn??? binaries.
>>
>>############
>>
>>Why? What's the point?
>>
>>Today, Walter has to really think things through beforehand. And do them PROPERLY in the compiler. All this slows down progress.
> 
> 
> Nice idea. My concern is that there'd not be time for Walter to do it. If this could be bazaar'd, maybe that'd succeed?

I always somehow thought that there'd be top level resistance
to Bazaar like things. But I had no reason, just felt like it.

There's such a lot of things that could be split and given
to interested crowds. "Somebody wanna look at the indexing
issue, and then we have this foo thing, that really could
use some thinking and proposals."

I remember Linus quite early started using more time reading
and commenting suggested kernel things, than actually writing
code himself. And the rest is history.