Holly Bostick wrote:

Matt Randolph schreef:
But does the Knoppix user's system have an administrator NOW?  I say
it does not.  It has been configured by an admin... heck, the OS was
installed to it's filesystem by an admin... but there is no admin looking over the shoulder of the Knoppix user.

Right.... so here's a real-world problem, from elsewhere on this list
("authorization failure when sending email")

Matthew Lee schreef:
I've tried every combination of kmail settings available, no joy. I've reemerged all the software that --depclean removed, no joy. I've reemerged kmail, no joy. I've reemerged ssmtp, no joy. However, I think ssmtp, or something associated with it is the problem. But what I haven't a clue. Is there another "simple" mail transfer agent I could try. I don't need anything fancy it's
just a laptop connected to the lab DHCP server.


Since this issue seems to revolve around programs also available to
Knoppix (and likely also being used under Knoppix), it's probably a
valid example.

So you've got a user who is unable to use a simple user function (send
email). In the proposed administratorless world, who is supposed to fix
this? The "invisible administrator" (who must exist, but is no longer
necessarily present).
Mr. Lee's problem is not that he cannot send email. It is that he cannot send email by the method he has chosen to use because he hasn't the knowledge necessary to make that method work. I assume he could probably resort to webmail in a pinch. If his distribution had provided those packages together with a wizard to bring the task of configuring them properly to within his grasp, he would not be having this problem. Is the task of producing such a wizard the responsibility of the Gentoo team? It would be only if he had paid them to provide such.

But Mr. Lee hasn't paid anyone to do this configuration for him. He has consented to serve the role of administrator for his laptop himself by choosing a non-commercial distribution without a tech support line. However, it sounds like he is administering his laptop in a reasonable fashion by first exhausting every idea he can come up with before turning to the community for help.

One might say that the admin is the person (or persons) that through knowledge and experience enable a system to perform what is required of it. Building systems that do not require physical interaction with administrators on a regular basis does not make the admin go away per se. It merely moves the admin FURTHER away. It may mean that the developers have assumed some of the roles an admin would have performed.

So if developers can produce software that actually is maintenance free (to the satisfaction of the enduser), what has happened to your administrator? Now it is the developer that has made the system work by virtue of their experience. If the authors of kmail and ssmtp can't or won't do that, there may be others who will. The paid people at Redhat and Linspire come to mind.

In the case of Knoppix, that's the Knoppix team or the Debian team, if
we're restricting ourselves purely to the packages involved. Is the user
supposed to download and install another "fixed" Knoppix disk in order
to be able to use KMail as they did last week? Or is the user to follow
the Debian protocol and not use the newer version of these programs
(meaning they wouldn't be available to Debian stable in the first place,
which of course, they probably aren't)?

If everything is supposed to "JustWork" and does not, someone must be at
fault. Who? The user is experiencing some unidentified conflict between
programs that worked together well last week. Is there any way for those
who are 'to blame' (development, packaging, some admin along the line)
to work in such a way that these conflicts never ever filter down to the
user? I say no, because we persist in making the conflicting
applications known to the user before all such conflicts are
identified and eliminated-- partly because development requires that
these errors filter down to the user to be identified in the first
place, as developers cannot test under all possible conditions.

Basically the limit of software technology is that we make it
immediately available to everyone as if it does not require
administration, but it is (almost) never so stable and intuitive that
this is in fact the case.

The solution would seem to be to either not make the software available
until it has been sufficiently tested so that it does "JustWork" under
all possible
conditions (which the trained greed of users will not allow), or teach
the user
that sometimes they may have to do something a bit more complicated than
just click 'Send' (which means that the user cannot be a pure user anymore).

I don't see any middle ground here, but maybe I'm missing something.

Holly
In the world of the cathedral, the middle ground is "both." Console video games cannot easily be patched after release, so the developers DO do extensive testing before the first product is shipped. Windows software users are required to configure their email clients in order to get their mail, but instructions are generally provided by their ISPs. In each case, those administrative tasks that can be performed remotely by experts are so performed. The enduser will always have to initially configure her email client to talk to her ISP just as the video gamer will always have to read the instructions to learn how to play a new game. [Though I think AOL users actually have all of the configuration done for them automatically. But then, AOL isn't strictly an ISP.]

The bazaar has settled on a "both" approach too, though. Linux distros have testing branches and stable branches in their package management schemes. Those users that can best be described as "endusers" should be encouraged to stay within the stable branch, while those users that can be better described as "administrators" should be encouraged to experiment with the testing branches. Even software in the stable branch will need to be configured properly to work, so knowledgeable people create Howtos and wizards for the benefit of the less knowledgeable. In each case, the novice receives the benefits of the experience of experts without having to be or to employ one of those experts. If the free help isn't helpful enough, the enduser needs to either get some skills or buy some personalized advice.

I think I've lost sight of just what it is we're discussing here. I'm not saying that there shouldn't be knowledgeable people serving administrative roles. I'm just saying that there is a place in the world for software that attempts to reduce or eliminate dependence on administrators. Will anyone ever be able to run an Enterprise email server without an admin? Undoubtedly not; someone needs to be on hand to ensure it keeps working and, if it stops working, to fix it RIGHT NOW. But should Linspire users be able to use email without having to know what a mail transfer agent is? I think so.

I think we have a difference of opinion on just what is meant by the word "admin." You seem to suggest that an admin is anyone whose interaction with an email client extends beyond merely clicking "send." I wouldn't call a person an admin simply because they may have installed a piece of software. Windows users install software all the time yet they seldom perform any of the other functions of an admin (if I had a nickel for each time I had to remove spyware that endusers had installed on their computers...). I don't think a person is an admin until they have taken significant responsibility for the care and upkeep of a system. The majority of personal Mac and Windows systems don't have administrators by that definition. Yet some of those systems often actually do "JustWork." Linux is actually an ideal operating system for those rank novices that have shirked their responsibilities as the de facto admins of their own computers. Properly configured, Linux is able to be incredibly stable. If I built a file server for my LAN and gave it a private IP, I could disconnect the monitor and keyboard and expect it to "JustWork." If it was plugged into a UPS and I had configured it well, it might "JustWork" continuously for many years until the hardware finally failed. Similarly, I believe a multi-function PC built on Linux, that would perform for years without a hiccup, is not an impossible target. If etc-update could be counted on to always do the right thing, that PC could even be running Gentoo. It's all a question of how numerous and complicated the demands are that are placed on the system. If a Linux box only has to do four things (say, email, the web, word processing, and solitaire), it is less likely but not unimaginable that one could be made to run for years without needing to be touched by an admin. If machines like that were placed in the hands of significant numbers of those people with such simple needs, users of proper Linux systems would reap benefits as well. Hardware manufacturers would have an incentive to find ways to release more Linux drivers. Commercial software developers (like game companies) would take Linux more seriously as a development target. Teenagers would discover that there was more to Linux than Mom and Dad's Linspire system and decide to join the rest of the club.

- Matt
--
gentoo-user@gentoo.org mailing list

Reply via email to