[Gllug] Cost of RedHat vs Ubuntu desktop support

[JLMS]

On Sat, Jul 11, 2009 at 1:29 PM, Nix<nix at esperi.org.uk> wrote:
> On 11 Jul 2009, JLMS told this:
>> You can standardize to 2 or 3 editors or varying complexity. You don't
>> need 10.
>
> Why restrict the users?
>
>> The reason for this is simple: every piece of software is an
>> additional problem waiting to happen (been there, done that, got the
>> t-shirt).
>
> I defy you to find a time that a *text editor* broke something else,
> except in ways that people can do with other text editors (e.g. the
> stupid Windows text editor habit of tabifying the whole file whenever
> you look at it: banning one text editor won't solve that disease because
> it's so regrettably common).


Are you seriously saying emacs can't brake things? There are people
that use it as a mini operating system (like the bozo that launched
200 ftp sessions simultaneously in a machine already heavily
oversubscribed).

In any case, that is not the point, I didn't claim that text editors
were corporate menaces, just that having many of each program (text
editors included) add to the amount of support required which is
wasteful.

>
>> The place for a software repository of the kind you are talking about
>> is a properly maintained software repository, not a user's home
>> directory.
>
> Yeah, right. That would mean giving me root, which would mean simply
> immense amounts of pointless bureaucracy. This is 'under the radar' so
> that it doesn't have to collide with the corporate bullshit which rains
> down from the big company overhead.

No, it wouldn't.

First of all the software repository does not need to be owned by
root, it can be owned by an account and group created specifically for
that purpose. In my experience in most cases when somebody was
claiming they needed a root account for something, they really didn't
(it was normally just a power trip, since having the root account was
considered a symbol of status or political power rather than a
technical need, or a lack of understanding of how to achieve a similar
end by different means), even if somebody really needs root access,
there are ways (sudo for example, although it is far from ideal) to
grant that access selectively.

You can automate the whole thing in a way that a developer drops a
software properly packaged in one place and an automatic process picks
it up, does some basic checks, and delivers a package available for
the relevant parties (i.e system administrators) in the other side.
The process can be something as simple as a cron job doing a copy, or
as complicated, as well, whatever you need.

The package can be installed from there in test servers, and finally
system administrators release the package into production (they do
have permission to do this of course, in most cases that means root
account, but it can mean many different things nowadays given the
existence of roles in Solaris, commercial software for this purpose,
or Secure Linux).

Notice that once you put a process like this in place (when there are
only 10 of you in the company), this scales quite well to the time
when you become a Google of a company.

Normally in places where bureaucracy is in place there are lots of
money to be lost, or serious consequences if mistakes are made. I also
used to think that some bureaucracy was pointless (those were the
years! low salaries and low expectations), until I had to spend
sleepless nights fixing something because somebody else decided that
the procedures in place where, well, pointless, and 3 or 4 hundred
people were relying on me to get their jobs done. .

My favourite is a chappy that decided to download top instead of using
the version provided internally. The problem was that top was not
using name services properly and was generating unintended DOS attacks
on some servers. Also a priceless "we are losing $10000000 an hour"
because somebody brought a piece of software that was DOSing our NTP
servers deserves special mention.

Like these, I have many stories, including, emacs, KDE, and many
others where we could track a very strange issue to an application
that was not allowed, and instead of improving systems or doing
regular work in  our servers, we were wasting our time tracking a
problem to a piece of software  in a desktop, that should have not
been there in the first place (yes, desktops can bring networks down,
and I am not even touching the Windows side of things, where people
equally forceful about  "their laptops"  have brought corporate
networks to their knees: viruses brought to the corporate network
performed DOS attacks, affecting UNIX/Linux  servers and desktops...).

Did you notice I don't even touch in the security side of things? I am
glad many of you trust blindly Sourceforge, I myself would quarantine
first any application, in case it wants to call home, and resources
available, I will have its code checked (install binaries not backed
by a company? Out of the question).

Honestly chaps, it is cavalier and frankly cowboyish to insist in the
do as I wish policy of software management. Any serious company should
not be working on that basis.
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug