[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: how to make Debian less fragile (long and philosophical)



[I'm cc'ing this to debian-devel because it has some relevance to the
conversation]

[and by the way, please read the whole thing before you start flaming me;
I will be agreeing with you at the bottom, and that's something...]

On Tue, 17 Aug 1999, Michael Stone wrote:

> > > to problems with libreadline, not libc. I have never seen an upgrade
> > > fail because of a libc problem.
> > 
> > I have, but it was six months ago and I don't remember the details. 
> 
> Not particularly helpful. :)

I know, but it does happen. My memory scrolls away after 2 weeks, so I
only have a vague recollection that it happened. I can't even be sure that
it was from an upgrade. But whatever. It's possible, it's worth
considering.

At work today, a big Solaris database server froze because its console was
connected to another Solaris box. Box #2 was taken down on purpose, but it
supposedly sent a Break to the DB. The DB halted. This was the main db for
a web site. Goes to show you, any small failure point is big and scary.

> Then get a clean copy off your rescue disk. Problem solved.
> 
> Why do you need to unpack a deb? I'm not after a list of things you
> can't do in a failure situation, I'm after a list of things you can't
> recover from using our toolbox.

Okay, I think I understand you now. These were where I went wrong:

  - I didn't know about sash
  - I didn't realize you were talking about _mounting_ a floppy, as
    opposed to booting off of it

Okay, my mistake, I take it all back, and I apologize if I let some
flamage through my attempt at reason.

However, I consider myself a fairly average guy, with a moderate amount of
sysadmin experience and 8 months or so running Debian machines. I didn't
know about sash, and I think lots of other people on this list don't, as
well (judging by some of the reasons for static binaries). So, as a way of
dealing with that issue, perhaps there are two things that could be done:

  - sash marked Important, so it's installed by default. It ask if it 
    should be root's shell, like it does now.
  - Debian needs some documentation on this stuff. This has been a problem
    with other issues (like the lack of an apt sources line to the
    security server in the security page, which is now fixed).

I think the first one is a good idea, but you might disagree. It's only
250K or so installed.

The second could be fixed with a Server Adminstrator's Guide or something
like that. Perhaps I'll do some work on it (assuming my writing skills
haven't totally atrophied with all the Perl I've been writing). I'll ask
about it on debian-doc, if anyone is still subscribed to it.

The mounted floppy thing is an intriguing idea. You could tell your BIOS
to not boot from floppy, then stick the floppy in, and you have a nice
little backup that's immune to hard drive corruption (if the kernel goes
crazy, it might corrupt other partitions in the same hard drive). This
allows you to still have remote recovery capabilities.

Another nice-to-have would be if sash had ftp capabilities, so you could
get to some other Debian box that isn't mangled. I'll ask the author about
it. Also, it would be nice if dpkg refused to remove packages that
Essential packages depend on.

> Installing by default!=essential. Essential means that someone can't
> uninstall it if they choose to. 

My mistake. Then just priority Standard or Important. Heck, if tetex can
be Standard, then sash should be at least that.


Reply to: