Perhaps I'm projecting my own experiences and mistakes but it seems like DJB's software and its comparative lack of success despite its superiority is a lesson in why it's more important to build consensus than it is to be right. No matter how correct you are or how good your solution is if you don't have the soft skills people might gladly go with a sub-par or even wrong solution.
I do think/hope that long term his software will win out though.
Well, I'd say that the software generally lacks packaging. I don't accept "read the uncommented hard-to-grok source" as documentation for any other piece of software. The programs are rock solid and great, but the source is nothing I would show a potential employer, and I am not a good enough C programmer to be able to rely on it for documentation.
I think of the lack of adoption as more of proof that in situations where hard facts rule, consensus rarely matches correctness. Social proofs, "hot technologies", tech fads, etc. are all demonstrations that much of the decision-making in both small and large organizations is much more "social" than technical.
This is why I am immediately skeptical of any SV venture that proclaims to be the new hotness in tech. They may be hot, but it's socially, not technically.
Except DJB's software was often not right (Qmail was a major offender against
SMTP protocol specification for a long time, until it finally got forgotten by
most of the sysadmins) or otherwise sub-par (djbdns lacked sensible zone
transfers, even non-incremental ones).
> Qmail was a major offender against SMTP protocol specification for a long time
No, it wasn't, ever. It is OK to offer an opinionated view of, e.g., qmail, but please stick to facts.
Some people used to argue that qmail's single recipient per envelope was some kind of violation. It never was, still isn't, so please don't repeat flame wars of the 1990's.
djbdns didn't provide zone transfers because they were a crappy way to achieve the desired effect and reduced the reliability of the service. Even at the time djbdns was published it was not best practice to enable zone transfers on production servers. No one today uses it for systems that have to be reliable and scalable.
> Qmail is an offender against smtp protocol specs? New one on me,
You must be using it for several years at most. In the old(ish) times, when
there were just Exim 3, Postfix, Qmail, and Sendmail, Qmail was getting quite
large flak from mail admins on various forums and newsgroups. Now, not so
much, but not because it got better (it didn't), but virtually nobody uses it
anymore.
> Zone transfers? Moving data is best done by other unix utilities, not monolithic programs like bind.
Oh yes, certainly. Let's do the same to replication in databases and directory
services (like LDAP).
For a time it was touch and go between qmail and all the other non-sendmail servers available at the time. qmail lost entirely because, I'm convinved, of its ridiculous licensing requirements.
DJB's code is good and would do the world good but he's a nut, which only hurts himself, his software and the computer ecosystem which lacks the lessons it would have taught.
I do think/hope that long term his software will win out though.