Re: posible bug in daemontools

It's going to be a good week, after all!

On 19 Mar 2010, at 15:05, Laurent Bercot wrote:
Any more questions?
> 
> Yes.

Splendid!

> Why the personal "fanboy" attacks? They're obviously not true. I maintain
> my own C programming libraries, *based* on DJB's code, but with so many
> additions, patches and rewrites that it's not DJB's code anymore - it's
> mine. If DJB was a god and wrote perfect code, I certainly wouldn't have
> to do this, would I ?

It doesn't have to be the code, though.  And "Fanboy" is never a personal attack.  It is simply the (probably well-meaning) audacity of telling somebody to think beauteous thoughts ...

> If you believe I'm telling Mark to stop thinking and join the cult,
> you're completely missing the point. I'm advising him to START thinking.

... the DJB way.

And, you are ***absolutely*** entitled to do that.  DJB gets a lot right, and I respect that.  But your dislike of the commonplace (yes, oft popular means crap) alternatives doesn't entitle you to tell people to stop thinking and start.  I don't like Debian because they modify things for no good reason, but they are absolutely guaranteed the freedom to do it.  If ever my software makes it to Debian and is then later bastardised, I'm going to make damn sure people know it's Debian's and not mine.  In the meantime, I look forward to Zinq as a new incarnation, and hope to see it birth a valid stock replacement that people can actually use that is secure, easy to install on 21st-century operating systems, reliable, etc, etc, etc.  It won't be far off from debian's current dbndns fork either.

> If he's going to distribute a djbdns-based package, is it so unreasonable
> to ask him to understand *why* djbdns was written like it is ? If he
> doesn't like the design decisions, nobody's preventing him from writing
> his own DNS package, that has nothing to do with djbdns whatsoever.
> 
There is nothing to prevent him doing whatever he wants with djbdns.  It's PD.  I wouldn't want DJB's reputation tarnished because of a bad, bad, bad fork, but we both know it won't happen because people use their brains and libel is only necessary when there are enough stupid people in the world who don't.  That is the ultimate remedy for somebody intent on fixing "Flaws" in software he clearly appreciates (based on the fact that he's, y'know, using and modifying it), and that's true regardless of whether or not you like his fork or the ideas embodied by it.

> Why do you think DJB's software is more disrespectful towards users
> than mainstream Unix software ?
> If it's only a question of "people are used to widely deployed tools,
> not following these habits is disrespectful to them", then I'm afraid
> Unix is very disrespectful and we should all be using Microsoft Windows.

Come now, that's blowing things out of proportion a bit much.  In reality, people are more tolerant than you think.  On this box, XMail (a completely monolithic mail server that is fast at threads) is running alongside Dovecot (split, but with a lovely admin-friendly front-end) and, of course, daemontools which not only starts and runs dnscache but also runs a job every 30 seconds because cron is too much trouble.  I can get on with them because I can accommodate them.  Yes, it's Debian, and yes, a part of that are the initscripts, but there's no law against admins actually knowing how their software works internally, true of all of the pieces on my machine.  So in reality there is less of a preference divide than you may think and less crisis about the right way of doing things.

> I don't know what you qualify as respect. To me, respecting the users
> means distributing quality software, with as few bugs as possible.
> Every security hole in a piece of software is HUGELY disrespectful to
> every sysadmin who trusted that software to work. Buggy software is
> insanely more disrespectful than software that doesn't fit in the mold.
> In my view, compared to other software, djbware is incredibly respectful.
> 
Software is respectful when it works.  Simple as that.  All of your points are valid insofar as that they must be true for the software to work.  By way of example, it is okay for OpenOffice to be a giant, heaving, monolithic pile of crap, because it's only useful when it's in one directory anyway, and it's okay for a mail server to be scattered throughout the file system because the only useful interface to it is a single command that is installed by the package manager that installed said mail server.

> I *never* pretended djbdns was perfect, or that it should not be patched.

Good.  Cooperative standards conformance is the reason I picked djbdns, and also the reason I take exception to it sometimes, for preferring behaviour that isn't always desirable.  Same for qmail, only worse in modern times.

> What I said is that it was thoughtfully designed. I'm not sure how anyone
> can disagree with this.
> 
I'm not really sure what this means.  Do you mean the source layout?  Do you mean the interfaces?

> And yeah, I do advocacy. Have a look at the GNU libtool code, or the
> OpenSSL API. And tell me they're not a pile of crap, I mean, disrespectful
> software.

I have been getting by chiefly because OpenSSL has enough documentation now that you can find your way around without too much ferreting about in the source, but I absolutely agree.  You can certainly go for something better if you choose, gnutls or nss, or build your OS detection/library building routines yourself, but I don't think I'm sufficiently disenchanted to do that.  Programmers suffer naturally.  We have reasons enough not to care that what we use isn't perfect, much as we would wish them to be.  It's a compromise we can live with, and it doesn't let us down (if it did, there'd be more alternatives).

A good weekend all!

Cheers,
Sabahattin