Upstart 1.0 development branch

Casey Dahlin cdahlin at
Thu May 7 15:31:25 BST 2009

Hash: SHA1

Scott James Remnant wrote:
> On Tue, 2009-05-05 at 15:21 -0400, Casey Dahlin wrote:
>>> In any Open Source project, it's quite normal for a developer to go off
>>> and do their own thing for a while before submitting or landing the
>>> code.  This is especially and triply true if they're doing some
>>> fundamental changes and don't actually know whether they'll work.
>> Its normal for a developer not to push things into mainline every
>> night. To drown their work in secrecy is quite out of the ordinary.
>> I've seen developers surface with code they hadn't mentioned before,
>> but I've /never/ heard of an open source developer refusing to show
>> code they were explicitly asked for.
> The only code I've refused to show, which is I'm sure what you're
> referring to, is the code to track forks and execs.
> There's a simple reason for this.
> I want this to be one of Upstart's "special sauce" features.  The
> ability to flawlessly supervise daemons, so you don't need to keep them
> in the foreground has been seriously lacking from any other replacement
> init daemon.
> This will _just_work_ in Upstart:
>   exec /sbin/syslogd
> If I release that code now, everyone else will know how I did it and
> then we'll see the same feature turn up in things like init-ng and
> einit.  Frankly, it took me a long time to figure it out, and I want to
> be first to have it.

That line of thinking is patently un-open source.

>>> I don't agree that having more developers work on a single piece of code
>>> somehow magically improves productively.  I'm following things down a
>>> particular rabbit hole that interests me.  If I have to stop and try and
>>> explain the changes I'm trying to make to others, so they can do it for
>>> me, and then review their work afterwards then things are going to take
>>> much longer.
>> Last time I mentioned I was working on implementing what had been
>> discussed you said you didn't want it, so I abandoned it (this was an
>> attempt at implementing things as you had envisioned them, not the
>> state machine).
> This never happened.
> You were working on your own state machine, which was crap.
> Worse, you were going around blogging about it as "THE new Upstart State
> Machine".

I'll take all the blame for that, but it is a separate issue.

After the state machine issue had been laid to rest I told you I was
working on implementing 0.10 as you had described it. You told me that
this was impossible, as you hadn't made up your mind on the way 0.10
should work. You also said that you were about to produce a design
document that would make these things firm.

I had dared to hope that this would be on a wiki where its formation
could be monitored and commented on. Instead the whole project simply
vanished under the sheets.

- --CJD
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Fedora -


More information about the upstart-devel mailing list