Re: A configuration tool for debian
Aaron Van Couwenberghe wrote:
> It seems to me that much effort is being reproduced on this particular
> subject. I agree completely with Wichert's spec -- as I recall it was
> adapted from an RFC he found somewhere, that had been through
> implementation. That kind of testing tends to solidify a spec.
No, it was all hashed out entirely on debian-devel last winter, and Wichert
put the spec through some 6 drafts to get it to where it is. No prior art
here.
> But now we have several projects:
>
> -Whatever Corel is doing (completely for non-interactivity purposes):
> Probably hacking individual packages, ie the whole distribution,
> to support a home-grown non-interactivity scheme
They have simply piped "yes ''" into the packages when installing. It fails
if just hitting enter all the way through makes a package install loop, and
locks up their installer. I saw this, showed them this, and have pointed
them at the source to debconf and they seem interested.
You forgot Stormix. They have hacked packages to be non-interactive, and
they're also interested in what I've been working on.
> -Doogie's dconfig
> What does this do? Reconnect dpkg's streams to a pipe going into a
> script that knows what dpkg will ask, and what it expects in response?
I think it is a modified version of the kernel config stuff for another
purpose. I wasn't aware it interfaced with dpkg.
> -Wichert and Ben's huge project -- dpkg2
> I would have to say this is the ultimate. It will take lots of work, but
> in theory it should solve the entire problem without cumbersome dpkg
> wrappers. It may even eliminate the need for apt (sorry Jason ;). IMO
> Corel would be better off helping this project out, although that would
> cause somewhat of a delay in their release ;)
Well when done it will need a configuration tool of some sort. I hope it'll
be the one in Wichert's spec, and I assume so, since he's working on both.
Note that Wichert's spec doesn't talk at all about how you interface with
dpkg.
> Do people begin such projects without forethought about collaboration?
Oftentimes yes. I hope I did a little better by just implementing a spec we
all hashed out together.
> Or do those who begin such projects underestimate their obligation to
> announce their work? Does this happen because there is no way anybody can
> discover who is already working to address problems like this? The
> community should be able to research progress of potential solutions.
Well I've underanounced what I'm working on because I don't think it is
quite up to the "plausable promise" state ESR mentions in the Cathederal and
the Bazaar as the stage a project has to be at before the open development
model works for it. In the menatime I've seen several people post on the
lists about similar projects and just told them what I already had.
--
see shy jo
Reply to: