I think you misunderstand -- or simply misstate -- the situation.
You are simply describing what being good at the job means. You are by no means describing what the worst case of having that job means. You describe a generic formula for knowing that you do a job well.
Do I understand my coworkers and communicate well with them? check
Do I have the skills necessary to function in my particluar role, and do I apply those skills properly? check
Do I understand our business and how my responsibilities affect it? check
The simple fact is that an admin has the opportunity to learn and apply a range of useful skills. So does a developer. So do the folks in marketing. And all of those people are responsible for some variation of your formula.
This does not mean that the sysadmin helps drive the company. If he wears many hats, maybe, but he's still the guy that gets called when there's a fire. Some of the advice given to the poor, benighted soul that asked slashdot for this advice was "get a job at a small company." Smaller companies have often failed to grasp the necessity of good IT practices. In such a situation, the senior man on staff for the job is often the only man on staff for the job. He gets to be his own piss boy even if he chooses your second choice.
And, effectively, the former english teacher doesn't get that last third of skill without going through hell first. There is no other way. Those who think otherwise just haven't had the shoe drop on them yet.
Sysadmin? Not me. Never me. No. Never. What? Money?
How much money?
I was swayed by money in my teens. I hacked, I played with gadgets, and I need the money to buy gadgets and pay for dinner. Sysadmin work payed the bills nicely. Why? Because there's always more sysadmin work than coding work, if only because most code is execrable filth used by illiterate idiots. Someone has to get paid to clean up the inevitable mess.
The problem is that the position of sysadmin is much like the position of janitor. You do a necessary job. You get insulted. No one cares about your other talents. You don't get paid anywhere near enough money for the sheer volume of shit (and sheer number of shits) you have to deal with every day. You quit. You get a job somewhere else and you never, ever, EVER mention that you were a sysadmin -- or even that you know what root is.
But eventually you start to care. You do good work as a hacker somewhere, but that somewhere has no sysadmin. You slip up; you're dumb enough to open you mouth or even stupid enough to fix something. Then you're a sysadmin again.
Here's your pager, here's your boss. There's an e250 talking to three gladiatior arrays holding five years of our data in a combination of oracle 7.mumble.foo and flat text files. Our lead developer is a seventeen year old in Pune and his assistant is a drunkard in Antwerp. They both have root access. We have no backups and no budget to buy tapes. Oh yeah, and the water pipes for the fire sprinklers have leaky valves right over that rack. Have fun. And I'd like an estimate of the costs of switching over to Exchange on my desk Monday.
How do people become sysadmins? Bad choices, bad mistakes, and punishment for sins a previous life. You don't want to do it. Asking this is like asking "How do I get my left testicle run over by a unimog?"
Trust me. Give up now. Go eat granola, smoke dope, sing hippie songs, and live under a tree somewhere. You'll be happier.
Lawrence lessig will also be appearing at the Commonwealth Club of California on Tuesday, February 20. This may be a more interesting appearance because the forum is not a traditional home of "cyber" issues. The Commonwealth Club is very much dedicated to all aspects of public affairs and has a very traditional (if staid) membership. I think it might be important to show policy groups such as this that these issues do matter in order to encourage more such events.
That Kamen has a habit of doing everything he says. The more ridiculous it is, the more likely he is to do it, do it well and make a great deal of money at it. The price range being talked about here makes it seem like his stirlings have been renamed.
They've even set up majordomo for a public list. Does that make anyone else nervous? Like maybe the confirmation isn't an email, but is instead two gentlemen with sunglasses, dark suits and menacing ties who come to you door?
"Someone, possibly you, subscribed to the mailing list 'selinux' at tycho.nsa.gov. Please confirm this action if you wish to subscribe."
I was about ready to dismiss this editorial until I read your response.
Frankly, I think you miss the point entirely.
I believe firmly that linux distributions (with the possible exception of
debian) are far too "big." This is not a reaction to being forced to change
the cd during the install! This is a response to the ever-growing task of
making linux boxen secure and manageable.
Let's look at the basic four machine installs that people do: workstation,
managed workstation, server and "other." A workstation is a machine meant
for users to have local processing, display and storage capacity in order to
accomplish certain tasks. What those tasks are can differ from software
devlopment to graphic design, or from surfing the web to network management.
A workstation is distinguished chiefly by being a machine that works in
conjunction with the services provided by other machines and networks.
Most workstations have no need for a local MTA, for instance. Nor do they
need bind. They almost certainly don't need DHCPD. No wu-imap, openldap,
ftpd, in.telnetd or apache are needed for most installs. Most systems don't
require xdm, kdm and gdm, nor do they require multiple compiler versions. In
fact, the desktop environments' coming on the distribution cd is not
something I perceive as an extra value over a windows install so much as an
extra annoyance. Sytem management and provisioning is an entirely different
logical problem space than application management. Most home users fall in
to this category. Most of them use windows (or macs) in part because the
distribution structure of those systems is easier to grasp.
Which brings us to the next category, managed workstations. These are the
workstations that operate in an environment with dedicated IS resources that
mangle^H^H^Hanage large numbers of similarly configured workstations. The
needs in this environment are very simlilar to that previously described but
the added need for batch management means that mechanisms must be in place
on the local workstation for the IS folk to initiate operations on every
unit. SSH, rsync and snmp are the basic building blocks here. The logical
separation of application and system management becomes even more important
in this environment than it was in the solo workstation envirnment.
The server environment is the one in which unix has traditionally been most at
home. This is a machine that provides resources that are shared among a group of
users. Here is your mail, web, dns and all the other stuff that leads to the
stories about "How Bob Saved His Company Seventeen Million Dollars with Only a
386, Two Orange Juice Cans and a Piece of String." Unfortunately, despite being
the hereditary stronghold for unix, the current distribution mechanism makes
even these tasks very difficult.
The traditional unix environment presents us with one box (server) doing many
tasks. This box is managed by either a group of people or by the local guru who
takes his time and carefully hand crafts the services and structure offered. The
modern environment, in this age of plenty, is to have many boxen
providing many services. The large distribution makes both types of
offering very complex.
In the first case, what with strange, artificial dependancies and wierd
assortments of code, our heroic admin must learn what is safe and what is
dangerous. He must know enough to lock down on extra cruft, et cetera,
yadda-yadda-ya. Again, this is now traditional. No one is shocked by it and
there are probably four articles a week for folks either faced with this task or
recovering from it. The distinguishing feature of this set is that there is no
logical separation between application/service management and system management.
While the larger distributions make the job more difficult they do attempt to
target the needs of this set.
The second case is more interesting, it is the case of the server farm or the
distributed network. Here, again, we see that the logical separation of service
management and system management becomes necessary. It is likely that our one
heroic admin has been replaced by many (not enough, of course) heroic admins who
all share responsiblity for the complete organism called the network. All of the
evils that the large distribution brings to one server have now been brought to
many (50? 100? 1000? Google?). The larger distributions -- those that don't know
what they want to be when they grow up -- make this task nearly impossible to
complete well.
The last case, the case of "other," has far too many possibilites to isolate.
This is the guy who uses his laptop as a roving server on an extruded subnet.
It's also the larval hacker who wants to learn how to write his own MUA while
running his mail server off his cable modem with dyndns. In essence, this
"other" case is the same as the initial server case in which one heroic admin
fine tunes his prize machine. This case is becoming ever less common once we
leave the living room, but the distributions do a great job targeting this.
In fact, the distributions do a great job of targeting the one-offs (anywhere we
find our heroic admin). At least, they do a good job initially. They do a
horrible job keeping those one-offs operating within the initial management
structure.
Compare this to NetBSD some day.I'll use Net as an example because it provides
what is arguably the leanest default install of any free operating system.
The NetBSD installation is barely acceptable for use out of the box and requires
a good deal of work to make a productive system. In the one-off world, there are
noticeably higher apparent start up costs (in terms of effort) for basic uses. I
say "apparent" costs because it is less likely that you will find yourself
removing software from a default Net install. Moreover, the incremental
costs of replicating these systems in a networked envirnmoent are close to nil.
Linux distributions are strange things that seem to want to cover all of the
bases for all of the possible users. They wind up falling terribly short and
causing those users to rework already existant solutions (as well as justifying
that effort to often-clueless management). A good way to judge their
effectiveness, however, is to see how they are used in the real world. Why don't
we ask the Slashdot folks to get some interviews with the MIS (desktop) boys
from VA Linux and the server engineers from Akamai?
It's not likely that any of these methods will make it "hard to ignore ads on the web." Sites that make use of these intrusive techniques will soon find themselves ignored.
The simple truth is that this method of advertising actively changes the state of a user's workplace. By changing the state of a user's workplace you force the user to undergo a series of context shifts. These will nearly eliminate the user's ability to actively view the original magnet content.
If joe user finds himself unable to focus on a source of content (due to the unbearable conditions in which it must be read), I don't think it an irrational leap to believe that he will simply stop visiting that site. It is very easy to ignore something with which you are never faced.
None of these methods will be serious problem in the short term. They are all hampered both by the medium and by the interface to that medium. The first important part to remember is that the web, unlike television, is not a serial medium. The browser interface to web content allows for a degree of parallelism in the tasks. Most web users, even the broadband sort, have simply grown used to reading in one window while another one loads. We tend to have many windows open at once. These ads, by there very nature, serialize the user experience, forcing the user to refocus from his chosen task of the moment by blocking his access to it. This is a serious flaw in design. (The browser popups are especially bad in this regard, since no current browser can accomplish that operation without a noticeable delay.)
This sort of product is the technological equivalent of a three-year-old's shouting "look at me!" As such, it will be avoided.
Over the last four years we've seen linux, and free Unix in general, reach heights of stability and performance that once seemed entirely unreasonable. In the last year, these free (to hell with OS!) operating environments have made equivalently great strides toward mainstream acceptance.
Dandy as that may be, all of tehse still have a ways to go. Xig addresses some of these problems in a traditional, and what seems to be an honorable way. They write good code.
So it's commericial! Big deal. Their only real competition in the intel-unix market comes from Xfree. Xfree, however, produces a product that is far less stable and speedy than AccelX.
Don't flame me for this one, guys. It's simple fact. At least, it's simple fact for now.
This ad, however, is disheartening. Xig's strengths as a company have always lain in their fair dealings, quick response and honest answers. Their products' strenghts have always been speed, stability and service -- and, sorry to be insulting here, but if you don't think service is an issue talk to me after you graduate from college. This ad, however, mentions none of the effort Xig has made over the last years. Rather, it slings mud at the equally honorable effort made by Xfree.
This ad, frankly, makes me believe that Xig is on its way to the grave. That would be a sad thing indeed. I can think of no reason other than blind fear for them to make a marketing move that is simply this stupid.
Anyway, I will continue to use their products where they are best suited (like working my Viper 770 Ultra under FreeBSD) until their product seems as poorly thought as their ad.
BTW: solution to the telnet in and kick X problem: hook a wyse terminal up to the serial port. Or would that be considered a kludge?:)
Slashdot Mirror
You are simply describing what being good at the job means. You are by no means describing what the worst case of having that job means. You describe a generic formula for knowing that you do a job well.
The simple fact is that an admin has the opportunity to learn and apply a range of useful skills. So does a developer. So do the folks in marketing. And all of those people are responsible for some variation of your formula.
This does not mean that the sysadmin helps drive the company. If he wears many hats, maybe, but he's still the guy that gets called when there's a fire. Some of the advice given to the poor, benighted soul that asked slashdot for this advice was "get a job at a small company." Smaller companies have often failed to grasp the necessity of good IT practices. In such a situation, the senior man on staff for the job is often the only man on staff for the job. He gets to be his own piss boy even if he chooses your second choice.
And, effectively, the former english teacher doesn't get that last third of skill without going through hell first. There is no other way. Those who think otherwise just haven't had the shoe drop on them yet.
How much money?
I was swayed by money in my teens. I hacked, I played with gadgets, and I need the money to buy gadgets and pay for dinner. Sysadmin work payed the bills nicely. Why? Because there's always more sysadmin work than coding work, if only because most code is execrable filth used by illiterate idiots. Someone has to get paid to clean up the inevitable mess.
The problem is that the position of sysadmin is much like the position of janitor. You do a necessary job. You get insulted. No one cares about your other talents. You don't get paid anywhere near enough money for the sheer volume of shit (and sheer number of shits) you have to deal with every day. You quit. You get a job somewhere else and you never, ever, EVER mention that you were a sysadmin -- or even that you know what root is.
But eventually you start to care. You do good work as a hacker somewhere, but that somewhere has no sysadmin. You slip up; you're dumb enough to open you mouth or even stupid enough to fix something. Then you're a sysadmin again.
Here's your pager, here's your boss. There's an e250 talking to three gladiatior arrays holding five years of our data in a combination of oracle 7.mumble.foo and flat text files. Our lead developer is a seventeen year old in Pune and his assistant is a drunkard in Antwerp. They both have root access. We have no backups and no budget to buy tapes. Oh yeah, and the water pipes for the fire sprinklers have leaky valves right over that rack. Have fun. And I'd like an estimate of the costs of switching over to Exchange on my desk Monday.
How do people become sysadmins? Bad choices, bad mistakes, and punishment for sins a previous life. You don't want to do it. Asking this is like asking "How do I get my left testicle run over by a unimog?"
Trust me. Give up now. Go eat granola, smoke dope, sing hippie songs, and live under a tree somewhere. You'll be happier.
Lawrence lessig will also be appearing at the Commonwealth Club of California on Tuesday, February 20. This may be a more interesting appearance because the forum is not a traditional home of "cyber" issues. The Commonwealth Club is very much dedicated to all aspects of public affairs and has a very traditional (if staid) membership. I think it might be important to show policy groups such as this that these issues do matter in order to encourage more such events.
That Kamen has a habit of doing everything he says. The more ridiculous it is, the more likely he is to do it, do it well and make a great deal of money at it. The price range being talked about here makes it seem like his stirlings have been renamed.
Those would indeed change the world. They would also be faced with an onslaught of opposition from the established companies in this area. People would plan cities around it.
If Kamen has really done it then I doubt that any of this is exageration.
"Someone, possibly you, subscribed to the mailing list 'selinux' at tycho.nsa.gov. Please confirm this action if you wish to subscribe."
I was about ready to dismiss this editorial until I read your response. Frankly, I think you miss the point entirely.
I believe firmly that linux distributions (with the possible exception of debian) are far too "big." This is not a reaction to being forced to change the cd during the install! This is a response to the ever-growing task of making linux boxen secure and manageable.
Let's look at the basic four machine installs that people do: workstation, managed workstation, server and "other." A workstation is a machine meant for users to have local processing, display and storage capacity in order to accomplish certain tasks. What those tasks are can differ from software devlopment to graphic design, or from surfing the web to network management. A workstation is distinguished chiefly by being a machine that works in conjunction with the services provided by other machines and networks. Most workstations have no need for a local MTA, for instance. Nor do they need bind. They almost certainly don't need DHCPD. No wu-imap, openldap, ftpd, in.telnetd or apache are needed for most installs. Most systems don't require xdm, kdm and gdm, nor do they require multiple compiler versions. In fact, the desktop environments' coming on the distribution cd is not something I perceive as an extra value over a windows install so much as an extra annoyance. Sytem management and provisioning is an entirely different logical problem space than application management. Most home users fall in to this category. Most of them use windows (or macs) in part because the distribution structure of those systems is easier to grasp.
Which brings us to the next category, managed workstations. These are the workstations that operate in an environment with dedicated IS resources that mangle^H^H^Hanage large numbers of similarly configured workstations. The needs in this environment are very simlilar to that previously described but the added need for batch management means that mechanisms must be in place on the local workstation for the IS folk to initiate operations on every unit. SSH, rsync and snmp are the basic building blocks here. The logical separation of application and system management becomes even more important in this environment than it was in the solo workstation envirnment.
The server environment is the one in which unix has traditionally been most at home. This is a machine that provides resources that are shared among a group of users. Here is your mail, web, dns and all the other stuff that leads to the stories about "How Bob Saved His Company Seventeen Million Dollars with Only a 386, Two Orange Juice Cans and a Piece of String." Unfortunately, despite being the hereditary stronghold for unix, the current distribution mechanism makes even these tasks very difficult.
The traditional unix environment presents us with one box (server) doing many tasks. This box is managed by either a group of people or by the local guru who takes his time and carefully hand crafts the services and structure offered. The modern environment, in this age of plenty, is to have many boxen providing many services. The large distribution makes both types of offering very complex.
In the first case, what with strange, artificial dependancies and wierd assortments of code, our heroic admin must learn what is safe and what is dangerous. He must know enough to lock down on extra cruft, et cetera, yadda-yadda-ya. Again, this is now traditional. No one is shocked by it and there are probably four articles a week for folks either faced with this task or recovering from it. The distinguishing feature of this set is that there is no logical separation between application/service management and system management. While the larger distributions make the job more difficult they do attempt to target the needs of this set.
The second case is more interesting, it is the case of the server farm or the distributed network. Here, again, we see that the logical separation of service management and system management becomes necessary. It is likely that our one heroic admin has been replaced by many (not enough, of course) heroic admins who all share responsiblity for the complete organism called the network. All of the evils that the large distribution brings to one server have now been brought to many (50? 100? 1000? Google?). The larger distributions -- those that don't know what they want to be when they grow up -- make this task nearly impossible to complete well.
The last case, the case of "other," has far too many possibilites to isolate. This is the guy who uses his laptop as a roving server on an extruded subnet. It's also the larval hacker who wants to learn how to write his own MUA while running his mail server off his cable modem with dyndns. In essence, this "other" case is the same as the initial server case in which one heroic admin fine tunes his prize machine. This case is becoming ever less common once we leave the living room, but the distributions do a great job targeting this.
In fact, the distributions do a great job of targeting the one-offs (anywhere we find our heroic admin). At least, they do a good job initially. They do a horrible job keeping those one-offs operating within the initial management structure.
Compare this to NetBSD some day.I'll use Net as an example because it provides what is arguably the leanest default install of any free operating system. The NetBSD installation is barely acceptable for use out of the box and requires a good deal of work to make a productive system. In the one-off world, there are noticeably higher apparent start up costs (in terms of effort) for basic uses. I say "apparent" costs because it is less likely that you will find yourself removing software from a default Net install. Moreover, the incremental costs of replicating these systems in a networked envirnmoent are close to nil.
Linux distributions are strange things that seem to want to cover all of the bases for all of the possible users. They wind up falling terribly short and causing those users to rework already existant solutions (as well as justifying that effort to often-clueless management). A good way to judge their effectiveness, however, is to see how they are used in the real world. Why don't we ask the Slashdot folks to get some interviews with the MIS (desktop) boys from VA Linux and the server engineers from Akamai?
It's not likely that any of these methods will make it "hard to ignore ads on the web." Sites that make use of these intrusive techniques will soon find themselves ignored.
The simple truth is that this method of advertising actively changes the state of a user's workplace. By changing the state of a user's workplace you force the user to undergo a series of context shifts. These will nearly eliminate the user's ability to actively view the original magnet content.
If joe user finds himself unable to focus on a source of content (due to the unbearable conditions in which it must be read), I don't think it an irrational leap to believe that he will simply stop visiting that site. It is very easy to ignore something with which you are never faced.
None of these methods will be serious problem in the short term. They are all hampered both by the medium and by the interface to that medium. The first important part to remember is that the web, unlike television, is not a serial medium. The browser interface to web content allows for a degree of parallelism in the tasks. Most web users, even the broadband sort, have simply grown used to reading in one window while another one loads. We tend to have many windows open at once. These ads, by there very nature, serialize the user experience, forcing the user to refocus from his chosen task of the moment by blocking his access to it. This is a serious flaw in design. (The browser popups are especially bad in this regard, since no current browser can accomplish that operation without a noticeable delay.)
This sort of product is the technological equivalent of a three-year-old's shouting "look at me!" As such, it will be avoided.
Over the last four years we've seen linux, and free Unix in general, reach heights of stability and performance that once seemed entirely unreasonable. In the last year, these free (to hell with OS!) operating environments have made equivalently great strides toward mainstream acceptance.
Dandy as that may be, all of tehse still have a ways to go. Xig addresses some of these problems in a traditional, and what seems to be an honorable way. They write good code.
So it's commericial! Big deal. Their only real competition in the intel-unix market comes from Xfree. Xfree, however, produces a product that is far less stable and speedy than AccelX.
Don't flame me for this one, guys. It's simple fact. At least, it's simple fact for now.
This ad, however, is disheartening. Xig's strengths as a company have always lain in their fair dealings, quick response and honest answers. Their products' strenghts have always been speed, stability and service -- and, sorry to be insulting here, but if you don't think service is an issue talk to me after you graduate from college. This ad, however, mentions none of the effort Xig has made over the last years. Rather, it slings mud at the equally honorable effort made by Xfree.
This ad, frankly, makes me believe that Xig is on its way to the grave. That would be a sad thing indeed. I can think of no reason other than blind fear for them to make a marketing move that is simply this stupid.
Anyway, I will continue to use their products where they are best suited (like working my Viper 770 Ultra under FreeBSD) until their product seems as poorly thought as their ad.
BTW: solution to the telnet in and kick X problem: hook a wyse terminal up to the serial port. Or would that be considered a kludge? :)