Re: JAVA and ADA JGNAT

[dvdeug@x8b4e53cd.dhcp.okstate.edu (David Starner)]

On Tue, 25 Jan 2000 23:06:06 GMT, Robert Dewar <robert_dewar@my-deja.com> wrote:
>Actually I think part of what goes on here is that ACT is
>*more* open than a lot of the Linux and GCC development.

No. ACT never releases beta versions to the general public, and 
does not have development lists open to the public, or even readable
to the public. I, as a developer, wouldn't work on GNAT because
of this. I, as a developer, have found and fixed a few minor bugs
in GCC, because I could.

>Take GCC, it is no secret that Cygnus does a LOT of GCC
>development that is closely held before being made public
>and often closely held for a long time. The same is true
>of course for Linux developments at Redhat. 

The way I heard it, any Cygnus-only developments have to be specially
marked in the internal tree, and they try to get new stuff in the 
public distribution as soon as it's paid for. Be that as it may,
Cygnus is not GCC's maintainer, and Red Hat not Linux's. For GCC, 
a steering commitee is in charge of it, and Cygnus' developments get
looked at the same way as anyone else's. No matter what they do inside
the company, the public GCC developments are public, and show no signs
of being a facade. Mark Mitchell, of CodeSourcery, is a prime example
that GCC is not controlled by Cygnus.

>It's never helpful to have testing jump too far ahead of
>development. In the case of JGNAT, the appropriate stage
>for the last couple of months has been to have a selected
>small number of beta testers kicking the tires.

"With enough eyes, all bugs become shallow." (Linus Torvald)
It's not improbable that the applet bug would have been fixed, had 
the right person checked out JGNAT. 

>The next step will be a general beta release, that corresponds
>to the sort of thing David Starner is talking about. 

No, I'm talking about development snapshots and stuff. 

>If David is saying that ALL developments should be made
>completely open day by day, all I can say is that I don't know
>of many open source or free software development projects that
>work that way, with the possible exception of GNOME (and a
>number of small scale projects).

Ironically, I think most small scale projects don't, because they
don't have the resources and interest to mount a CVS server and
mailing list. But the mainstream of many projects work that way, GCC
and Linux especially. The main development source is out there to 
study - sure, someone may have been developing this over here, and 
someone else this that hasn't been merged in yet, but the main flow
is there. And the head developers usually aren't happy with this side
developments, because they make merging in terrible.

>I definitely think that would
>not be helpful to GNAT users. Yes, it might be fun for a few
>enthusiasts and hobbyists, but the confusion of having lots
>and lots of versions of GNAT around, most of them being works
>in progress that were non-functional

Why is this a worry? It doesn't seem to have been a major problem
for Linux and GCC. ACT also has the advantage of a smaller community
that isn't as likely to abuse the privilage.

I'm not espoucing anything that radical, just basically the
whole Cathedral and Bazaar stuff.

-- 
David Starner - dstarner98@aasaa.ofe.org
If you wish to strive for peace of soul then believe; 
if you wish to be a devotee of truth, then inquire.
   -- Friedrich Nietzsche