LISA’14, the Large Installation System Administration conference, was held in Seattle last week. I had the opportunity to give a talk on Server Management – if you are interested, the slides are available here.
One of the questions caught me completely off guard: “Aren’t you afraid that you are making system management too simple and that people won’t learn how to really manage Linux? They will just learn a few simple commands and not go any further. Today, they have to learn how Linux works and how to solve problems. OpenLMI will leave them unprepared.”
Wow… Where to start?
Thinking about this further, it could happen. In fact it will happen! Many people are looking for the quickest fix to a problem – a common way of working is to Google what you need, find something that looks like it should work, try a quick cut and paste, and move on.
OpenLMI is designed to support this. The LMI CLI is task oriented, simple, and easy to use. All you really need to use the LMI CLI is “LMI help”. The LMIShell scripts are designed to do useful work, to be easy to read, and to be modified for specific tasks.
If someone is simply looking for a way to perform a specific task, use it, and move on the the next problem, OpenLMI is a good way to go. You can use OpenLMI at a shallow level, even use it to avoid having to learn how Linux really works.
On the other hand, OpenLMI can also be used to ease into a deep knowledge of Linux: Start with the LMI CLI and use it to perform tasks. Move into LMIShell and start using and developing scripts. From there it is straightforward to develop custom automation tools. You have several ways to dive deeper into Linux administration, perhaps even developing custom OpenLMI Providers.
I would suggest that OpenLMI makes Linux more approachable. Some people will only use OpenLMI, and will never go deeper – if they can do what they need to do, this seems like a reasonable approach. Some people will use OpenLMI as a tool and and entry point to mastering Linux administration; this is great.
I don’t believe everyone needs to master Linux to use it. Consider the car analogy: All some people want to do is drive a car – automatic transmissions are perfect for them. Some people want to be able to do light repairs such as oil changes. Some want to rebuild engines and repair major subsystems of the car. And some people want to design the eight speed computer controlled automatic transmissions that are part of the integrated drive train of modern cars!
What do you think? Do we face a real risk of making Linux “too easy”, or should we try to make Linux more approachable?
techponder 
Well.. The idea behind IT is that computers should simplify our lifes.
So, if the the tools are enough reliable, why not use them?
There are too schools of thought:
One says to get the immediate task done as quickly as possible and move on to the next task.
The other says that people need to have sufficient in-depth mastery of a subject to be able to do original work and to solve difficult problems – if you don’t raise people to this level of mastery you are setting them up for failure.
I can see valid points on both sides of the argument, and would like to know what other people think.
A question at the Red Hat BOF at LISA was “Who is pushing this agenda on making Linux so easy to use? Why are you trying to put us out of a job?”
I’m rather surprised to see this coming up in the Linux community. Did other people agree, or was it just the one person raising the question?
My reaction is “do you really want to spend your day setting up storage and networks, installing software, and fixing passwords, or would you like to free up some time to work on interesting things?”
There was a lot of head nodding so it wasn’t just one person. I think the problem is that their boss says that this is great because they can replace their $75k Linux admin with a $50k MSCE admin like they had replaced their $150k Unix admin with a $75k Linux admin. Because neither they nor the boss know what interesting things they could be working on because they have been draining swamps for so long that all they know is alligators and pythons versus new stuff.
OK, this is great feedback. The easy way out would be to say “if your job can be replaced that easily you need to be adding more value”, but that seems rather facile.
Were there any thoughts about what could be done?
(For the record, I’m ignoring your python pun!)
I missed that I punned. I was just watching a show in the hotel about anacondas taking over top predator spots in Florida/Georgia swamps. I think the issue is that when 99% of your job is firefighting and you are paid bonus money to be a firefighter… installing automated sprinklers doesn’t look like a great deal. Figuring out how to add more value can also be a hard problem to see when 99% of the time is pager duty. [Well even at 60% it is hard to see.]
The IT personnel are not the end users. The end users are those who they provide service for. And what is easier for them is not always better for the end users. The everyday reality of modern enterprise is when I ask IT department to do something, which I know is perferctly easy to do – like change a line in config file, or add a simple script into some procedure which is designed to execute them. But quite often I get a response “sorry we cannot do this, we don’t have such button in our management interface”. By providing “easier” way for IT personnel, solution vendors effectively let them provide worse service for end users.
I am not blaming anyone for being personally reponsible for this, just the car analogy does not work here.