Re: posible bug in daemontools

It's been a bad week, and I'm not going to make any friends.  So ...

In your fanboy rant, this part needs the most attention:

On 19 Mar 2010, at 08:34, Laurent Bercot wrote:
Sure, but existing standards fail to address a problem that slashpackage
> does address. Part of the DJB philosophy is: if the existing standards
> don't meet your technical requirements, don't follow them. Do not settle
> for mediocrity, but design the APIs you need, for a technically better
> product. And that very philosophy gave you dns_tranmit_start() over
> gethostbyname().

It also gave us qmail.
http://www.dt.e-technik.uni-dortmund.de/~ma/qmail-bugs.html

Whatever you may say about his software's "Technical excellence", it is marked by an unbelievable and frank disrespect of others, and in several clear cases is WRONG.  Demanding respect of others under such circumstances is therefore somewhat dangerous in the presence of flammable liquids.  Any more questions?

> Now do you understand how going ahead and forking djbware (and I
> mean *distributing* software, I don't care what people do at home, of
> course) without the utmost prudence can be seen as disrespectful?

No.  He can do whatever he damn well pleases, and for whomever it may so benefit.  He does not need to first gain the appeal, love and personal favours of DJB or his minion of faithful followers.

Now, don't misunderstand.  I'm no fan of autoconf either.  I think Daemontools is Dan's best and most useful work, and appreciate some of the decisions he has made.  The Unix philosophy, the simplicity, is best illustrated by his software.  I use dnscache, and might even have used qmail, because it met my needs (although nowadays Courier provides a much better alternative with the same design goals).  I love Dan's writings, and for a lot of purposes I imagine his software is still suitable, if only because most of the compromises he makes are acceptable or at any rate non-destructive to those who can't or won't read RFCs.  I may even be inspired, at least in part, in my own software, especially by his 2007 paper on a decade of Qmail.  And, of course, changes made to system software should be done with caution.

But Dan is not special, a god, somewhat for whom respect is given but not taken.  If you had any doubts, check out some of the mechanisms used by Apple, Courier, Postfix, Dovecot or numerous others.  All the good bits without the obsolete rubbish and seniority complexes that prevent the software from gaining any kind of useful deployment in the 21st century.

Cheers,
Sabahattin