[nSLUG] [OT] Looking for a DNS secondary partner

Ian Campbell ian at slu.ms
Mon May 4 18:52:30 ADT 2009


On Mon, May 04, 2009 at 05:45:02PM -0300, Daniel Morrison wrote:
> 2009/5/4 Ian Campbell <ian at slu.ms>:
> > On Mon, May 04, 2009 at 03:17:08PM -0300, D G Teed wrote:
> >> On Mon, May 4, 2009 at 2:51 PM, Ian Campbell <ian at slu.ms> wrote:
> >> > On Mon, May 04, 2009 at 10:52:19AM -0300, D G Teed wrote:
> >> I do think it is a professional failing.  yum is used in a professional
> >> product or two, and given its critical role, I'd think they only allow
> >> skilled coders to tackle its maintenance.
> >
> > Two words: debian openssl.
> 
> Two words do not a cogent argument make. What?

It's used in professional products, it plays a critical role, and
you'd think they'd only allow skilled coders to tackle its
maintenance.

It's not uncommon that critical software has bugs, and I don't think
it's a binary choice: that either Kurt Roeckx is an unskilled coder or
that he's a bad person.

> > Even OpenSSH has had its share of bugs. Mistakes happen, even for
> > skilled coders.
> 
> Earlier you were saying that a high moral standard (e.g. not engaging
> in murder) is not required to write good code. Now you're saying that
> even good coders make mistakes. Conclusion: having low moral standards
> does not exempt you from making mistakes, even if you're a good coder.
> Am I following correctly? :)

You are.

The point I was attempting to make was sort of the inverse, that
having high moral standards (or not) is utterly, totally and
completely irrelevant. That everybody makes mistakes and that whether
you help old ladies across the street or push them into traffic has
zero impact on the quality of your code.

> > To be fair, from a developer's perspective a python backtrace is
> > probably far more useful than "Error fooing Bar"
> 
> But it sucks from a user's point of view.
> 
> > I know nothing about your code, but I'm going to go out on a limb and
> > say a backtrace would be more useful than a generic error message
> 
> You added the word 'generic'.

I did, because a useful error message is also known as a backtrace.

Which of the following is more useful for a bug report?

1) SEGV
2) "Error opening critical file"
3) "Error opening critical file in FILENAME.EXT on line LINENUMBER"
4) A backtrace

> > there too. You may want to handle it more gracefully than yum does,
> > but... yum doesn't need to keep running, it has the luxury of being
> > able to exit abruptly if it wants.
> 
> yum dosomething withmypackage
> [abrupt exit]
> 
> Oh great, now what?

Try again.

yum dosomething withmypackage
[oh noes error, backtrace follows]
[abrupt exit]

> Yum is production code used at the core of distributions' package
> management. I don't think it has the luxury to break unexpectedly...
> and yet it does.
>
> > I guess I don't understand what your complaint is. Does yum blindly
> > assume things without checking them (obviously bad), or is your
> > complaint that it dumps an unfriendly-looking flood of messages on
> > users when it fails?
> 
> It was stated that yum blindly assumes a file exists before attempting
> to open it, and then crashes. That's what I remember reading earlier,
> anyhow...

My mistake, I confused that with his segue into his own code about
assuming a connection succeeded etc.

> > > > What I never understood was why ext3 enjoys the popularity it does.
> Stephen Gregory said:
> > > Ext3 has stability and reliability. Two nice features when you are
> > > dealing with data.
> 
> > XFS has been around almost as long as Ext2, performs better, and has
> > other nice features not found in Ext2 (like safe online dump, online
> > defrag, and a couple features like guaranteed IO rate that I don't
> > think were ever ported to Linux... still, they exist.)
> 
> I remember trying to setup XFS on a RAID I was creating at home, back
> when I still thought SCSI was cool (maybe 7-8 years ago). It was a
> nightmare trying to understand the complex xfs tools, not to mention
> getting xfs support properly compiled into the kernel. All kinds of
> third-party patches.

If you tried it before it was in mainline, you were asking for
trouble. I'm not aware of any significant warts after SGI switched to
Linux for their products... it was in mainline in 2.4, I forget what
year that was.

I'm not sure what complexity there is in the xfstools (xfsdump and
xfs_fsr are both pretty straightforward) ... and in any case they all
have man pages.

> Fundamentally, ext3 enjoys the popularity it does for the same reason
> Windows, MSIE, and FAT32 do. Because they're the default. All Linux
> kernels since forever supported ext2, and ext3 is backwards-compatible
> with ext2. All distributions include the tools. You can boot any Linux
> rescue disk from any distribution and it will be able to read your
> partitions. There's drivers to read ext2 from Windows (and likely
> other OSs) too.

When has "it's the default because it's the default" ever been a good
reason? I mean I know this discussion isn't going to change that or
anything, I'm just saying ;)

Can't think of any distro that doesn't have support for JFS, XFS,
reiser, or ext off the top of my head.

> Stability, reliability, availability. And the fact that it's the de
> facto "Linux standard" means that it will still be supported tomorrow.
> That's why it's popular.
> 
> ReiserFS was surrounded in controversy (due to Hans Reiser's ego) long
> before he committed any well-publicized criminal acts. That's the main
> reason I never got around to testing it. I got the impression that if
> it broke, I could pay for commercial support, or risk a tongue-lashing
> in some forum. Not so with ext2/3/4. I've met Stephen Tweedie and Ted
> T'so, and they're nice guys! So I use ext3 and get warm fuzzies.
> Simple.

What controversy? LKML is for big boys, he was just one overgrown ego
among many. See also Ingo Molnare, Con Kolivas, Alan Cox... there was
a lot of bickering over Reiser4, but that was mainly because it broke
from Linux's VFS subsystem and reimplemented a lot of stuff, would
have required more effort to maintain because there was duplicated
functionality etc.

No tongue lashing? See Ted T'so's response to the ext4 posix behaviour
drama.

> PS: one can (and many have) said the same things about OpenBSD: safer
> than Linux, better, etc. But people like Linus; he makes them feel
> good. Not so for Theo.

I'm not holding OBSD up as an example for anything other than that
assholes can write good code.



More information about the nSLUG mailing list