Re: using CVS and Re: [Gzz] Paperbot status

[Tuomas Lukka]

On Tue, May 13, 2003 at 09:10:01PM +0300, Tuukka Hastrup wrote:
> On Tue, 13 May 2003, Tuomas Lukka wrote:
> > On Tue, May 13, 2003 at 05:49:43PM +0300, Tuukka Hastrup wrote:
> > > First problem comes with files being in some specific places and owned by 
> > > some specific user who doesn't have write access to CVS. So I'd have to 
> > > do 
> > > anonymous checkout and symlink the files. And every time I twid 
> > > something, 
> > > I'd need to commit and update. Here one more problem is that the system 
> > > spans several machines and I'd need to update several times.
> > 
> > So problem one is laziness? ;)
> 
> You could call it laziness :-) I could call the situation impractical. 
> Could you handle a situation where you'd always need to commit before you 
> could test the effect of your modification?

An alternative: "make publish" = copy the current cvs repo onto the 
right locations. All you need is the right password to go from you
to having right to write the files &c.

> > > Another thing was that the scripts contain configuration variables which 
> > > are better unpublished.
> > 
> > Ok, valid as well.
> 
> You could call this one laziness as well ;-) But seriously, in the 
> prototyping phase, the separation would've brought in otherwise unneeded 
> complexity.
> 
> I have to admit I probably went too long before targeting the addition to 
> CVS, as CVS undeniably has strong points for it. What do you think, do all 
> the deployment scripts of STORM be in CVS, even though they are of the 
> type "cd to some directory, cvs update and call some make target"? I think 
> they're not that important, and the time could be better spent in 
> documenting the interfaces so that anyone could come up with such scripts 
> easily. I mean documentation such as storm/README's "Our software can:" 
> and "Usage".

Hmm, I'd strongly prefer if CVS contained *everything* that's needed if 
all our machines' hard drives suddenly crashed.

Leaving out one script can easily add a day to the downtime (you don't
think it would do, right now, but think when you happen to be away,
and the machines crash and someone not as familiar needs to redo things..
a day can be an *optimistic* estimate).

        Tuomas