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

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

I guess it comes down to this:

  1. I think it costs next to nothing since I don't believe there
     is any performance benefit in a dynamic "cat"

  2. Most systems will benefit from extra reliability; and it would
     help Debian build a reputation as being reliable 

  3. Reliability is an attitude.

Many people have argued that there is some performance benefit in 
a dynamic "cat", "sh", or "ls", but I think there is none. There is
a win in having a dynamic "emacs", something that sits in memory 
all day and is fairly large. There is no win in something like "ls"
which is small, exists in memory for only a brief moment, and
overall just isn't loaded 99.9% of the time. There is a disk size 
factor, but lots of packages are bigger--it could be "important"
and you could choose NOT to install it, but it would be the 
default to have it. Also, I think the amount of diskspace we 
are talking about is just going to get more trivial over time,
not less.

Most systems really would benefit. I have given all kinds of examples
of critical systems--and those are the ones which will make you the 
most proud when they have Debian up and running--but everyone benefits
when a machine can safely be fixed up from remote. If it saves you 
from walking into the office when your upgrade-from-home goes sour, 
well it bought you something, even if only of small value.

And to be blunt, my biggest concern is that if reliability gets 
brushed off on this list because "those who don't crave static 
linking never miss it". Well if reliability features get traded 
out because 99.9% of Debian users don't care about it, then we 
might as well just admit we're building a desktop OS, because 
this is just going to be the first of many tradeoffs. 

I can't fight the tide--I don't have the energy to track down all
the differfent special little things I have to do to make Debian
useful to me. I honestly will just run something else on machines
where it matters.

I raised it as a policy issue because I want to know, I really 
want to know, what I can expect from Debian in terms of reliability:
is it important and critical and core? Or is it something that's nice
to have, so long as it doesn't get in the way of more important things,
like saving 0.1% of your RAM, and having a 20% smaller base.tgz?

I am not stuck on the particular solution that I proposed. I am 
stuck on the belief that the current situation is intolerable, and 
something has to be done about it--whether it is my proposal, or 
something else that is just as effective.

I will be happy if the problems I've raised get dealt with in 
an effective and convincing way.


On Tue, Aug 17, 1999 at 07:23:29PM -0500, Nathan E Norman wrote:
> On Tue, 17 Aug 1999, Justin Wells wrote:
> [ snip ]
>  : Debian has got a lot of the key things in place, such as a good 
>  : organization, a policy, a package manager with many strong points; 
>  : but I feel it is still a little lacking on the reliability side.
> I don't follow your argument.  You seem to be saying "Because high
> relieability is important to me and a few others, all Debian installs
> should use statically linked binaries in certain locations by default".
> You are saying this by raising this as a policy issue.
> Can't a "high-reliability" package be created which contains
> statically-linked binaries?  I imagine such a package would need a
> policy exemption for static linking, but who cares - that is the point
> after all.
> This way, people who crave static linking get it.  Those who don't,
> never miss it.
> Idea, or am I talking out of my ass?
> --
> Nathan Norman
> MidcoNet  410 South Phillips Avenue  Sioux Falls, SD
> mailto:finn@midco.net           http://www.midco.net
> finger finn@home.midco.net for PGP Key: (0xA33B86E9)
> -- 
> To UNSUBSCRIBE, email to debian-devel-request@lists.debian.org
> with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

Reply to: