Linux Desktop Deployment Postmortems?

duffbeer703 asks: "My employer runs alot of desktop and laptop computers -- something in the neighborhood of 40,000 PCs. Currently they are all Windows 2000 & XP managed by Active Directory and other big, complicated enterprise management tools, all of which can support Linux in one form or another. I'm looking for ways of making Linux (and maybe Unix or even Apple desktops) an option as we replace or add PCs. The problem is, most of the resources that you find online about deploying Linux focuses on server environment, and the articles that I do find about desktop Linux focus on standalone developer workstations, the IBM conversion to Linux (which doesn't seem to have happened) or things like LTSP, that won't integrate well with our infrastructure. Is anyone out there successfully using Linux for regular users? How did it go, and how did your IT and user communities adapt to the new kid on the block?"

Comments

Ubuntu?

[abscondment]

This article was posted a little while ago about a user who used Ubuntu in a completly MS environment without his boss noticing for a few months. (linked article from the story)

My experience with it is that it's one of the most mature Desktop distributions, coming complete with most of the tools one would need to perform most jobs. Easy install, and you can use Syntaptic/apt-get for upgrades and additional installation since it's Debian based. You should check it out.

Re:Ubuntu?

[GigsVT]

Sigh.

First the Red Hat zealots came for me, and I said nothing because I didn't want to run a commercialized distro
Then the Gentoo zealots came for me, and I said nothing because I didn't want to compile everything
Then the OS X zealots came for me, and I said nothing because I won't pay for overpriced hardware
Finally the Ubuntu zealots came for me, and everyone was so sick of offtopic zealotry that no one spoke up at all.

Re:Ubuntu?

[HolyCrapSCOsux]

Here's my story. I worked in a callcenter. VERY M$-centric. They were using MSAccess to handle all their supervisory overhead. (write-ups, time tracking,HR stuff) and it was horrible. the damn thing crashed EVERY DAY.
This is a server story--->
I got permission from my boss (who was not in the IT department) to build a proof of concept web based replacement for 200 users.

system:
Compaq Armada 7400 Laptop
PII 300
64 MB RAM
Slack 10

Now, Everything was going well for months. All the supervisors were happy. The system was operating flawlessly. Then one overly ambitious assmonkey decied that he could curry favor with the site manager by filing a complaint against my "going against IT policy by having an unauthorised server".
The IT department was fully aware of the server being on the network. Obviously. They knew I wasn't going to break their infrastructure. They merely looked the other way, as it filled a need.
Nonetheless, the project was scrapped. Everybody who used the system got pissed at the guy who filed the complaint because it tripled their workload. That guy resigned. Ha Ha. But it still terminated the project.

Moral of the story: if you want to try to help people and "increase productivity" get top-down approval of the project.

Re:Ubuntu?

[Dan Ost]

Sounds like you're perfectly positioned to get top-down approval now considering you've already demonstrated a clear need, know what kind of improvement you can expect, and have the support of the people who will be using the system.

Try it again and let us know how it goes.

Re:Ubuntu?

[Robert The Coward]

Old data. The client is now free and included with both Fedora Core 3 and 4. It doesn't work well at all it is slow and crashes way to easilly. If they turn on IMAP support the problem goes away thought. As I use Exchange All day long using IMAP without any problems at all.

Head first

[saskboy]

IT's probably best to dive into a Linux or any OS migration for users head first, all at once, so everyone in the office has identical migration problems and can assist each other if the official tech support is busy. It's like the choice between staying with paper, or going with computers, that businesses had to make in the '70s, '80s, or '90s. There will be some people who would never bother to learn unless they are tossed into it kicking.

Size matters

[Professional Heckler]

Your employer runs a pretty hefty workstation. Although I have worked for, or known people that made similiar switches the scale was not even close. So it worked pretty well as the community was close-nit and excited about the change.
In your case though, there will be more disruption, not everyone wants to use linux... Id suggest just inserting the new computers in one department, preferably one where the employees are already interested in linux. I would also suggest taking a workgroup poll to get interior feedback interest as well.

prof

Not here, either

[NineNine]

Just so that nobody thinks that nobody is reading this thread... No Linux deployments at my company. I don't think that we'll look at Linux again for at least a few more years. None of our important apps work on Linux, and we have no Linux expertise in our small company.

Guitar Strings

[jag7720]

Take a look at the Ernie Ball guitar string company. They made the switch several years ago. It is only 300 +/- people but they did it cause they got hit with being out of compliance with M$

Read Rockin' on without Microsoft

Re:Guitar Strings

[everphilski]

*whispers* (welcome to slashdot)

-everphilski-

Re:Guitar Strings

[Parity]

Don't be stupid. Out of compliance means you have more installations than you can -show- that you have licenses for. It doesn't mean that you actually have pirated software, but that you didn't keep all the right certificates of authenticity in a secure place. If your process wasn't tightly controlled, you can easily be out of compliance when somebody cleans out a closet full of 'old junk that nobody needs' (aka, all those still shrink-wrapped windows manuals).

Furthermore, being out of compliance is not a criminal matter, it's a civil one, so even if they were in actual violation and not merely in a state of poor record-keeping, they still wouldn't be 'criminals'.

(Yeah, yeah, IHBT, I dunno when I'll learn...)

Re:Guitar Strings

[pla]

I find it interesting that criminals are touted as a Linux success story.

A "few dozen" unlicensed apps on a network with 300 people shouldn't have warranted BSA-imposed pariahdom. A small fine and forcing them to fix their licensure status, yes. Disgracing them on the evening news and in an ad campaign, no.

Although it might count as blasphemy to say as much on Slashdot, Microsoft, of all companies, understands that, and except in really extreme situations will usually work with a company to get them in compliance, for NO fine (even offering a discount to "help them out" in some cases). The BSA, on the other hand... Absolute pure evil. It amazes me that anyone would allow them on-site without a warrant and a police escort.


You also have to wonder what "unlicensed" means, in context... For only a "few dozen" installations, does that mean they accidentally exceeded their number of VLKs? Shareware that had expired without buying the full version? Random programs that employees had brought in from home that the company didn't even know about? "Out-of-upgrade-path" upgrades (meaning, for example, that you can't take an OEM Win95 machine to XP via an upgrade copy - but you can upgrade it to NT4, which you can then upgrade to XP)? Plenty of situations to consider before calling them outright criminals. Oh, by the way, you need to fix that tail light, sir - I'll let you off with a $65K fine this time.


Yes, you can say that none of those situations should have occurred. But welcome to the real world, where even the most diligent IT department can't catch everything.

Re:Guitar Strings

[Coryoth]

Although it might count as blasphemy to say as much on Slashdot, Microsoft, of all companies, understands that, and except in really extreme situations will usually work with a company to get them in compliance, for NO fine (even offering a discount to "help them out" in some cases). The BSA, on the other hand... Absolute pure evil. It amazes me that anyone would allow them on-site without a warrant and a police escort.

That would be the "good cop, bad cop" approach. You need a decent stick to beat people with (the BSA) but you also need to be able to present yourself as the "good guy" trying help the person out...

Jedidiah.

Ghost et al.

[meisenst]

I was able, at some point a few years back, to produce a Ghost image with Red Hat, OpenOffice, and a login model that used my office's Windows infrastructure to authenticate users automatically. It worked very well. I used it on several test PCs and was able to boot them up, ghost them, and have them come up connected and ready to use.

It was fairly straightforward to set things up with simple additions to /etc/skel. The only real kneebiter was the fact that the vast majority of the office required Outlook, and for some reason (I don't recall what) Evolution wouldn't quite cut it. I seem to recall problems with lookups in the Active Directory using Evolution, but for all I know that's been fixed by now.

I ran this thing on my PC for months before my employer even noticed. I used VMware for my Windows needs (as I was a network administrator, I needed to run some troubleshooting in Windows for user support) and Samba for all of my day-to-day shares and printing. In the end, the only reason anyone knew what I was running was that I was sick one day, and someone tried to sit at my desk, with very small amounts of success.

Now if only I'd kept a copy when I was let go!

You know

[sjvn]

You could just give Red Hat or Novell a call and either one will be more than happy to give you their dog-and-pony show for their desktop offerings. I mean, they do do this kind of thing for a living these days.

Do you have must-keep Windows apps? Try CrossOver Office

http://www.eweek.com/article2/0,1895,1886920,00.as p

or

Verasora/Win4Lin

http://www.versora.com/

I've used and deployed them all in small businesses with AD management, and they've all worked. There's no reason why they wouldn't work in larger businesses. After all, as IBM and Oracle are showing, they already do.

Steven

Re:You know

[DrXym]

You could just give Red Hat or Novell a call and either one will be more than happy to give you their dog-and-pony show for their desktop offerings. I mean, they do do this kind of thing for a living these days.

The funny thing is that for companies who do this for a living, where are the migration tools? Microsoft provide a shit load of migration tools for moving away from competing products and even old versions of their own products. Why doesn't Novell or Red Hat?

Think how attractive it would be if RH / Novell could back up and install over an NT server. If it could replicate all file & printer shares and take the existing NT PDC / Active Directory settings and implement the Linux equivalent. Think how damned cool it would be if they even converted the NTFS partition to ext3 while they were at it so data remained in-place. Or if they printed out a handy checklist of things to be done after installation to complete the conversion. In other words make conversion from NT / W2K / XP as simple and painless as possible.

This wouldn't be a magic bullet, but even if it meant that 60% of installs could be automated, it would be a very, very good thing for Linux.

Rationale to a company...

[HerculesMO]

There's a few reasons why an IS department won't roll out Linux into an Active Directory environment.

First, is that they cannot control the desktop using policy. This is the biggest selling point of using Windows in a workgroup domain, and especially to manage as many servers and end users as they have. Active Directory, while not perfect, is awesome in its capabilities -- all stolen mostly from Novell's NDS :)

Next, is expertise. Why would you introduce something into an environment that nobody really knows how to use? Your executives aren't 100% sure but they know 100% that they need to hire staff that can take on Linux servers/desktops and supporting them. That means paying a premium for that labor, and it's not necessary when you can get Windows guys on the cheap.

Lastly -- companies are hesistant to change. Financial companies in particular go with the mantra, if it works, don't touch it. You will see lots of these smaller shops on NT 4 still because to them... it works. Larger corporations that have to meet with SOX compliancy issues are forced into upgrading. That's what happened where I work.

Anyways.. best of luck trying to introduce Linux into your environment. I am going to say that you will crash and burn trying, because a company that large doesn't likely have a *need* for Linux. And if's not a necessity, a good business decision is not to let it happen. Again the mantra, if it ain't broke...don't fix it.

Re:Rationale to a company...

[amightywind]

Lastly -- companies are hesistant to change. Financial companies in particular go with the mantra, if it works, don't touch it. You will see lots of these smaller shops on NT 4 still because to them... it works. Larger corporations that have to meet with SOX compliancy issues are forced into upgrading. That's what happened where I work.

Interesting posting, but you are forgeting an even more powerful force in corporate culture - corporations are always looking to cut costs. That $10M and growing annual check to M$ will eventually overcome the cultural inertia. It has been slow going for about 8 years, but it is happening.

Re:Rationale to a company...

[molarmass192]

Why would a company arbitarily cut an annual check to Microsoft?

Annual support contract, most companies have them.

WTF

[drownie]

You know I read your rant/article about gnome some time ago, posting it into random stories as comment doesn't make it any better ... http://linuxtoday.com/news_story.php3?ltsn=2005-11 -04-018-26-OP-SS-NV-0089

Desktop Linux in the Enterprise

[John the Kiwi]

I've been trying to get Linux instaslled on the desktop for a few of my customers, but had problems finding a suitable model for deployment. Say what you will about Microsoft (and here most people do) but their deployment tools are pretty good. All of my new deployments utilise RIS (Remote Install Services) which greatly reduces client installation times.

Roaming Profiles and publishing applications via Active Directory also greatly reduces on site time. Workstations can be restored without anyone technical being required on site at all.

I've looked and looked and haven't been able to find any resources for doing similar tasks with Linux based desktops. The closest I've come up with is to use custom built CD Rom desktop OSes, but these are much slower than using a workstation with the OS installed on a local hard drive.

I'm sure it can be done, perhaps by remotely mounting common application and /home folders to a central server. But I've never seen any Howto's or even descriptions of anyone having done this in the enterprise before. Not to say it hasn't been done, just that noone's written how it's done (that I've been able to find).

Not much help I know, but it shows why my company is still an MS shop.

John the Kiwi

Re:Desktop Linux in the Enterprise

[pla]

All of my new deployments utilise RIS (Remote Install Services) which greatly reduces client installation times.
Roaming Profiles and publishing applications via Active Directory also greatly reduces on site time.

RIS helps greatly under Windows because you can't just install everything you want and then image the drive (unless you plan to put it on 100% identical hardware, and even then you need to sysprep it). With Linux, you can do an absolute base install in about thirty seconds more than it takes to format the HDD, then just do a normal copy from a CD (or networked) image to a live system. Or if you trust the drive, you can even skip the install, and just dd an image directly onto the HDD.

As for roaming profiles - Just put home directories on a network share. Simple as that.

Not to say that one should try to force Linux onto otherwise unwilling companies and users... But most of the reasons I've heard not to switch simply don't hold true.



I'm sure it can be done, perhaps by remotely mounting common application and /home folders to a central server.

For apps, include them in the base image you put on each machine. For home dirs, you apparently already understand the easy and obvious solution.

Re:Desktop Linux in the Enterprise

[Dan Ost]

Can I do a base install of Linux in 30 seconds like you say? What technologies would I use?

Here's how I've done it in the past:

0. make tarballs of the machine where I built the system image (with users and everything). Make them once, use them forever.

At the machine I want to install on:
1. boot off a CD or network image
2. create boot and root partitions (~10 seconds)
3. install the bootloader (~10 seconds)
4. fire off a script that untars the boot tarball onto the boot partition, untars the root tarball onto the root partition, ejects the CD, and then reboots the machine.
5. walk away.
6. come back later to adjust the BIOS if necessary.

How do you make sure the kernel is compiled with all of the appropriate drivers?

If each machine is the same as the machine you imaged, then there's nothing to worry about. If not, then build a kernel that has everything you're likely to need built as a module or use a kernel from Suse or Redhat or Knoppix where they've already done this for you.

The reason that you don't see much documentation on this subject is that there really isn't much to it. If you understand how a linux system boots and know how to use common tools like tar and ssh/wget/curl/whatever, you can build this type of thing from scratch faster than you can look it up online.

Head over to the gentoo wiki and see if they've got an article that describes this sort of thing. If they don't, request one. They are amazingly responsive to requests and, while the resulting guide will be aimed at gentoo users, once you see how they've done it, adapting it to another distro should be simple.

We're Switching Because... It's Cool...?

[RobotRunAmok]

Let me state that I love Linux, and I am fortunate enough to be able to use it for my work.

In the past I've been responsible for switching a small company over (circa 150 desktops) from -- what was it now? -- DOS to WIN 3.1, or WIN 3.1 to WIN 95, I forget, I've burned it from my memory. And it was a nightmare. Not cuz it was Windows, cuz we were switching, period. Accounting gave us hell ("what are the cost benefits again?"), users gave me hell ("Time is Money, Y'Know!"), and Super Senior Mgt tweaked me more than once ("If you weren't switching us to this, um, upgrade thing, what is it that you would be doing, hmm?"). Learned an AWFUL lot about wacky boutique Accounting-Inventory-Shipping-Graphics-YouNameIt programs that all ran lovely on the OLD system but had to be bludgeoned into submission on the new.

Not saying you should not upgrade. Not saying Linux is not an upgrade from what you're using (not saying it IS, either; you really need to examine the apps). Just saying that you really need to look at this upgrade from every direction short of Sunday before you dive into the change. There's a large, cold room reserved in the House of Pain for Linux Evangelists who push their companies to make The Switch without having a whole pond worth of ducks in a row.

Good Luck, Bud, and God speed! And better you than me.

Re:Microsoft writing Slashdot titles?

[amliebsch]

I can't believe PHB's are using "postmortem!" The term they are looking for is "After-Action Report", or "AAR" in mil-speak. Tell them that using military terms makes them sound bold and dynamic, while using medical pathology terms makes them sound weak and dying.

Disaster

[TedCheshireAcad]

I used to work at a private high school in the Northeast. You can probably figure out what one by looking at my user name. Anyhow, we (read: I) tried a rollout of Linux on our file servers and routers. Here's what happened:

The Linux file server worked beautifully. We had a simple NT4 domain, setting up Samba with proper permissions was easy. It was easy to administer, very reliable, and fast.

The Linux router(s) worked well, too. I had a nice collection of scripts run with cron that would turn off internet access to the dorms at a specified time, and then turn it back on in the morning (remember: this was a high school).

I was even in the process of developing a grading system with the LAMP stack, since at the time, teachers did their grading manually, and often complained about it.

Everything was running beautifully for months, until politics entered the game. Some higher-ups bought software without consulting the IT department (me and one other guy) that of course only ran on Windows. They also decided that we were going to go with FileMaker for a grade database, that was maintained by some high-price consultant. In the end, they wanted everything to be Windows for some reason or another (misinformed about how Open Source works, you know, the whole deal). My wonderful little Linux environment disappeared, and eventually, so did I.

Moral of the story: technical challenges aside, your project can always be torpedoed by someone who is self-important and more powerful than you.

Re:Disaster

[stevey]

I've had similar things happen to me in the past, more than once.

The most common experience I've had is working with a small company with approximately 50 staff on site, and a few remote. The backend is entirely Linux based, Exim for Email, Apache for the webserver, Samba for roaming home directories, etc. (Each desktop user will typically have an Windows 2000/XP installation, some brave types [like myself] might run Linux, and no Microsoft servers at all.)

Fast forward a year or two and the company gets bought out.

The next thing you know the entirely open, working, stable, and proven backend is replaced en masse by a Microsoft solution - to make it identical to the software that the parent company has been using.

Having recently been through this for the third time I'm quite cynical. It is almost painful to see a company suddenly lock themselves into paying for upgrades and still losing services which were available previously.

Not to mention switching from nice open POP3+Imap to "improved" installations such as Exchange, or Lotus Notes is enough to make grown men weep.

Re:Disaster

[itomato]

"Moral of the story: technical challenges aside, your project can always be torpedoed by someone who is self-important and more powerful than you."

Amen, brother.

Self-important twats have spoiled the flowering of several potentially beautiful projects.

The problem is not many people dig a paradigm shift, and Windows to F/OSS is definately one of those.

First Switch your Servers

[mgpeter]

I have successfully deployed GNU/Linux networks, both servers and workstations. If you are at all serious about deploying a large amount of GNU/Linux Workstations the first thing you should do is replace the Windows Servers.

It is much easier integrating a Unix type workstation if you use Unix type servers. It is trivial to have nfs mounted /home directories, especially when you use LDAP for the User Database. If you attempt to deploy Unix type workstations in a Windows Network enviornment you will ALWAYS be fighting with the servers.

In your environment it'll be tough

[Sycraft-fu]

If it's all Windows centric including backend and management, it'll be tough to add. Here we are a hybrid Windows/Solaris and are adding Linux. The way we do it is LDAP on Solaris for the backend. Sun has a product that syncs the AD to LDAP, and we are currently working with the Linux systems to get them all working. They use LDAP just fine, but we are having difficulty with our automounts and other such things.

If you want to do it in your the thing to look for would be a way to sync Linux with the AD. I don't have any experience in this area so I'm afraid I can't help, but Samba might be a place to start. I understand it works in Windows 2000 domains now. At any rate what you want is to design a solution such that the existing management tools will work more or less seemlessly with the Linux workstations. That means they need to get their account information from the AD, map the Windows file shares (Linux does that fine now) use the Windows printers CUPS has no problem with that) and so on.

You will probably need a Linux server that's the go-between and you might have to do some custom development work. However, I'm sure it's doable. Remember though, to sell it you need ot make your solution work with the existing one. If you demand a bunch of changes, you'll just get shut down. However if you make it integrate nice, it's much easier to push as an alternative. Ultimately a more platform-neutral back end would probably be good, but with infastructure that large, you can't start there because the cost will be enough to make everyone say no.

PRobably what you should do is just get permission to start experimenting. Get a Linux desktop and server up and running under your control and then start investigating what it's going to take to get some integration going on. Worst case, it doesn't work out, and you get some Linux experience out of it.

Re:Not One

[Lumpy]

most small businesses freak when they see a real accounting package. Peachtree and Quickbooks are NOT real accounting packages but toy packages for the business owner that does not know accounting.

Real systems like Champion controller and sage and Cougar mountian or even Excalibur.

Those that are still using the toy packages the likes of Quickbooks really do not want powerful, they want braindead and to pay a service fee to get the hard stuff done.

but that is the difference between buying a $395-$595 toy at compusa or staples and a $1500-$6000 accounting suite from a professional.

kick it up a notch

[__aaitqo8496]

okay, i really hate the subject line (and emeril's show) but here goes:

i work in a very small environment... say roughly 25 employees and at least that many desktops with about 20 servers. i've been pushing to move away from being a microsoft shop. luckily, the guy before me was also very pro-Best Solution (note i didn't say pro-linux or anti-microsoft) and set up a number of linux servers.

i have taken hold and attempted ot push the idea of linux desktop solution for people that don't need windows (i.e. sales people). i actually set up a second box for myself before deploying a test box for a sales person. being a ubuntu user for 3 releases now, i choose it for it's polish, shine, and my comfort level. my experiences have been mostly good. anytime anyone needs a package, i just grab it from apt-get (or find a repo first if need be). i can take care of the whole box via ssh and never have to bother the user. it works GREAT except for a few small problems in a windows network:

1. setting up active directory authentication is a PAIN. it's not hard, but time-consuming and requires a lot of manual tweaking (see my request for an automated tool)
2. evolution-exchange connector is horribly in need of work. the basics work, but it's not fast or efficient - or stable. it gets the job done, albeit not eloquently
3. (i belive the following is a problem with nautilus, but idk) when accessing a shared windows folder, authentication gives a prompt for credentials, but it doesn't matter when you put here. the second prompt for credentials is the important one. in fact, you cannot get the first box to go away unless you click cancel
4. sudo & AD groups. for the life of me i can't figure out how to get sudo to recognize %domain\linuxadmins as a valid group. `groups` shows me as being part of it, but it's almost as if sudo doesn't like the slash. i've tried escaping it, and tried it without the domain to no avail. ideally, i'd like to set up a group to allow certain users to perform updates when ubuntu notifies them stuff is in need of updateing.

my gripes aren't HUGE, but they're annoying to me. of course i haven't touched on management needed for a 20,000 pc environment (pushed software & updates), so ymmv

Re:kick it up a notch

[NateTech]

In a managed environment, the users shouldn't be doing that anyway. It should be a crontab installed on their machines, immutable, that checks to see if you've added anything to the repository nightly.

That's the part you seem to be missing. Users shouldn't need or have to update their own machines in a managed environment -- you should be in control of that from start to finish.

Re:Out of compliance?

[CoderBob]

http://www.infoworld.com/articles/hn/xml/02/11/27/ 021127hnerniball.html?s=IDGNS

News story from the event. The article is light on the details, and at one point refers to "pirated copies" while at another refers to "more installations than licenses".

Having seen both many a time in a corporate environment, this is not always a company decision- users are to blame on occasion as well.

The reason for the shift matters, but the fact that they shifted successfully says a lot, especially to smaller organizations that might not be able to afford enough licenses. If those style shops start switching over to avoid being out of compliance, things could start to get real interesting.

by request only

[LodCrappo]

I work as a consultant for smaller companies. Although I use linux on my desktops and am quite happy with it, I wouldn't recommend any of my clients try to deploy it on the desktops for normal users unless there is some very compelling reason to do so. I've yet to come across such a situation, but I guess cost, performance and/or security might be reasons in some cases.

On the other hand I do have some clients where certain individuals have requested linux, and allowing them to run it has not caused any problems other than the obvious compatibilty issues that may apply. These individuals are linux savy and can generally deal with their own problems. Management does not want to spend extra money to support a second platform, and they understand this.

I guess the point I'm trying to make is that if you are considering rolling out some linux or even apple desktops, I would be careful to only migrate people that really want them and understand the consequences (and are able to deal with their own problems for the most part). Otherwise you're going to be incurring extra costs that probably outweight any licensing money you save. That usually doesn't go over well and will generate a negative attitude from management towards linux.

As for workstation management tools, there are solutions from Redhat and Novell and probably others, and IBM has some tools too. I don't have much experience with any of them, but again it is probably an extra cost and what would the point be? What is the boss going to like about this whole idea? Sometimes I think linux fans push too hard or don't fully evaluate the situation and actually reduce the opportunities they might have to use linux where it would really be a great solution.

Re:Not One

[NineNine]

Right tool for the right job. For many small companies (mine is around $1 mil/year), an expensive accounting package isn't worth the price. My $400 copy of Quickbooks does everything that I need it to do right now. Do you have a good business reason that I need a mid-range accounting package right now?

Success stories

[slashdot_commentator]

I find it interesting that criminals are touted as a Linux success story.

Not if you live on a continent full of "criminals" with success stories.

Tried Sabayon? (Useful for Gnome envs + profiles)

[GingerDog]

For a Gnome based desktop, Sabayon appears to be about the best thing I've found yet that allows you to create "profiles" for different users.

I don't think it's anywhere as good as what I've heard group policy to be, but it's a start in the right direction. I've found it to be quite buggy and it took me a couple of days to get the desktop _as_I_wanted_it_.

(See http://www.codepoets.co.uk/sabayon_creating_linux_ desktop_profiles which may be of some use as feedback/info)

DG

Can't Switch For Switching's Sake

[zoomba]

Especially in a company with that many desktops. When talking about a migration to Linux in a large environment like that means a bunch of things:

1. What do you do with ANY of the custom apps used on the desktop. Most large companies have at least a few apps their internal developers built for them, and I'll bet they weren't built with cross-platform use in mind. Sure, it may work for now in WINE, but what about when it throws a weird error? What about when a new feature is needed? Recoding the app isn't really an option for most places.

2. Time to fire and rehire your desktop support staff! And any IT group that is directly tied to desktop products, cause you're doing a complete 180 degree switch on them. You can argue that anyone worth paying should already know Linux, but the reality is a lot of people in IT are tied to MS, because that's what their company has bet the farm on. You would probably have to either rehire or retrain most of the desktop support group.

3. Your options are RedHat, or SuSE. A company that big is only going to switch if they can buy Linux from a vendor with the chops to support a large organization. Mom & Pop Linux Support Inc isn't going to be taken seriously since they may be in business today, but might not be tomorrow. Business wants a large company backing a product so they have someone to go back to when something goes really wrong.

4. Retraining Costs. Sure, there's adjustment when moving users from Windows Version X to Windows Version Y, but generally the user experience remains fairly consistent. Moving to Linux, unless you reskin it to look exactly like Windows and hide away anything that would hint that it wasn't Windows is going to require significant user retraining. Then there's all the new apps that they'll have to learn to use. You'll lost a LOT of time and money here.

5. What's the real benefit? Yeah, Microsoft is evil, vendor lock-in, security vulnerabilities blah blah blah and so on. But honestly, does Linux provide a real business value? Does it save money in the long run? Does it make the work easier to do? Don't answer these questions as techno-geeks who are already biased, look at it from a semi-objective standpoint.

I don't think you can make an effective case to begin the switch-over of 40,000 desktops to linux, even in much of a phased approach. Best you can probably hope for are a few pockets of Linux users within IT. The average user would probably never even get whif of its existance.

Re:Can't Switch For Switching's Sake

[Shakes268]

Best post I've ready all day. Every single one of these is a valid point and something Linux has to work to alleviate concerns about if it is going to penetrate the desktop in large organizations. Linux pretty much has to become commercialized overall, like RedHat for companies to get the "support" safety blanket. They aren't going to listen to some guy in sandals wearing a t-shirt with a penguin on it telling them he can handle all 20k PCs. Not only do you have to fire your support staff and re-hire people who know Linux but you have to look at the cost in regards to the new people you hire. Linux people are going to be more expensive and any cost you probably save in licensing you might end up paying in personnel. 95% of the users in corporate America don't "need" Linux. Most don't even know what it is! Most own Windows PCs at home so they can transfer work back and forth easily as well as have consistency. Personally, I have never worked in a company - as a FTE or as a consultant that wrote desktop software with cross platform availability in mind. The mentality is always do it fast, do it as cheap as possible with as little problems as possible. If you bring up,"In 10 years you might want to move to Linux so we're going to write everything in perl and java - it will take us a little longer, dev rates will be a little more expensive and your current infrastructure doesn't support some of the things we need to do to make it work - but you will thank us in 10 years" That just won't be acceptable to the people signing off on the projects, paying the checks and running the show.

Re:Can't Switch For Switching's Sake

[Nik13]

Great post. I had said something similar in a previous topic a while ago as well.

Most people I hear talking about "this Linux thing" at work (most of the time users) don't know a thing about it, except that it's free (and windows isn't) and they somehow think that would pretty much drop our overall IT costs to 0$.

Truth is... Even if you don't look at the money figures (may or may not be better - let's just focus on the other issues first which sometimes are more of a concern)

We have DOZENS of in-house apps (working on a new and faily big SOA n-tier app in C# / .Net FW 2.0 app currently). Simple ones, huge ones, and in different languages, some "legacy" stuff too. There's LOTS of odd stuff users have to run on their desktops, and re-coding/replacing all these alone would most likely cost more than the windows to linux switch would save and could perhaps be more disruptive too.

Not to mention the TONS of of "not-quite-apps": big access databases (with hundreds of forms and tables - a huge mess that would be near impossible to port to anything in the same century), visual foxpro stuff, excel documents with VBA (lots of those), etc. Lots of it sucks badly, but we just can't take it away from them.

There's also the common commercial 3rd party apps which are only available for windows... Be it for HR people, finances, legal dept, doing statistics, planning and charting, web design, GIS, etc etc. That would be HUNDREDS of small apps to find replacements for (and there very well may be none). This may cost lots of money again.

All our infrastructure is around MS stuff. Lots of stuff relies on SQL Server. We use Active Directory and Exchange Server. Our intranet (thousands of pages and over a hundred web apps)? All ASP.Net (in C#; and some "classic ASP" stuff left) on IIS 6. Suggesting to PHP-ize everything, replace IIS for Apache, and to ditch our (already licensed & paid for) SQL Servers for Oracle or DB2 ($$$!)? Force all our coders into Java and/or PHP? I don't know, but there just isn't a good solution here really.

This list is like never ending. You just can't possibly re-code/replace all this, or force everybody to use a terminal server for half of the stuff they need everyday... In the end perhaps costing more money and creating a lot of trouble...

And like you mentionned, retraining costs, of users and admins/support people (IF you don't just have to hire completely different people instead).

We have ~60000 users, and we're making the switch from 2k to XP right now and it's very painless. Licensing costs? Well, I'm sure it's lots of money, but you gotta keep the big picture in mind too... 60000 users with an avg pay of 60000$ (my guess - may be completely off), a heavily-discounted XP upgrade license around 100$-some dollars every few years for a windows upgrade seems like not that much money (I know I cost them more PER DAY than a copy of XP costs, and that's the first upgrade in years). Our new PCs even come with XP too (no upgrade costs).

A switch to linux on the desktop (we do have a few linux servers)? Don't think it'll happen here anytime soon. Not because there's anything wrong with linux, just that it's a change to a completely different solution where everything works differently/isn't compatible with our current proprietary stuff, so it just won't happen. Any other non-Windows OS would pretty much have the same results (drastic changes required everywhere). Linux is great and all, but it's not always the one and only solution to everything.

How to be successful at migrations

[einhverfr]

First, be patient. I don't think the IBM migration is as dead as it appears. Most of the commercial migrations I have seen take 2-3 years to accomplish assuming that a fair amount of resources are thrown at the problem. If you want a smoother transition, I would suggest planning for 4-5 years. This timeframe should allow you to rewrite all your inhouse applications to support Linux if necessary

The first step is to identify those workstations that have the simplest requirements and/or the users who are most interested in switching. Start there and migrate a few stations at a time. Don't be afraid to rollback to WIndows for a while when you need to. Try to use Wine and other technologies to make the transition easier. I think that this is still where IBM is.

The second step is to do an analysis of what has/has not worked in this step and then look for the next group of workstations to migrate. Wash, rinse, repeat until you run out of shampoo.

Once you have a fairly established set of Linux workstations, I would suggest investing in infrastructure. Look at things like OpenAFS, X11 application servers, and the like. For desktops you can create a computing network that looks conceptually sort of like a SAN and is very easy to maintain (read up on Project Athena). This requires more care with laptops because of mobility requirements,but if you are careful about which applications you put on the laptop and which ones you run over the network, you should have few issues.

Hint: You can put an X server on the Windows systems to give them access to your X11 app servers, and therefore not immediately require everyone to rn Linux to gain access to certain applications.

Re:How to be successful at migrations

[sgtrock]

That's just the US, though. Off the top of my head; Europe has the Extremadura (sp?) project in Spain, the postal service in Poland, and Munich. South America has OSS projects going in Peru and Argentina. In Asia,there's a relatively large OSS effort going on in China.

So. Just because us North Americans haven't moved off the dime doesn't mean that the rest of the world is sitting still. :)

Re:How to be successful at migrations

[einhverfr]

First, if you set it up properly, there is no reason that upgrading your system should be harder on Linux than on Windows. THere are great tools like Yum, apt-get, etc that can be scheduled, and you can push out configuration files via scp and shell scripts if you like.

Secondly, you have far less work interruption from updates on Linux than with Windows. With tools like apt-get or yum you could indeed upgrade the distro without taking it down for the upgrade or booting the user off his/her applications (depending on the work, the user might notice a performance hit though). This is part of the bit about investing in infrastructure. With the right work, you can make Linux *far* easier to manage than Windows on the desktop (yes, I know what a GPO is). Secondly... with the right infrastructure of AFS, LDAP (or Hesiod), Kerberos, and X11, you can have something that requires very little work to maintain.

Re:How to be successful at migrations

[einhverfr]

Can you even get support for a 5 year old version of Linux?

I cannot find any information on how long RHEL versions are supported with Red Hat's support engineers. However, they maintains update support for (and expect you to be using it for up to) seven years, so I would assume that the answer to your question is probably "yes."

MacWindows

[wheatwilliams]

If you want to integrate Mac OS X computers into your existing Windows server infrastructure, be sure to check out http://macwindows.com/

This site is dedicated to enabling Mac OS X computers to coexist in the enterprise environment.

Linux enthusiasts will never catch on

[couch_warrior]

A year or two ago, I orchestrated an enterprise upgrade from Win 9x desktops with Banyan servers to WinXP with Win2K servers.
You would not believe how scared and panicky the users got. During the physical migration, users were given 4 hours of training on the changes from Win9X to WinXP. Then immediately went back to their desk to a newly converted workstation. It hardly helped at all. The shape of the MS Office icons changed, we got dozens of calls from users who said we had "taken away" MS Office. One department had their shared drive change from the P to the Q drive letter. Even after telling them verbally three times in class, and following it up with email, we still got dozens of calls from users who said their documents had been "deleted". We even got calls from people complaining that their spyware was missing! And some of them were PhD's.

Humans in general are dumb, easily panicked sheep. They fear and loathe change as if it were physically painful. They don't like Windows - in fact it is one of the favorite topics of water-cooler derision. But they would rather run their nuts through a clothes-wringer than have their toolbar move to the top of their screen.

Linux enthusiasts are generally highly intelligent malcontents. People who desire frequent chaotic change because it soothes the agony of their ADHD induced boredom. They love having to follow up the latest installation of Fedaro by trying to figure out where to download a multi-media player from because *someone* got pissy with the old player's authors and left it out of the distro. Tweaking /etc/initab to customize the services running at a particular run level is a diverting amusement rather than an odious burden.

And Linux enthusiasts suffer from a terrible conceit, believeing that the rest of the world "wants" to be like them, but just doesn't know how. So if they can make the Linux desktop look 75% the same as windows, then they can lure the sheep in for a closer look. The implicit assumption being that once a sheep gets a good look at the "freedom" offered by vi and shell scripting, they'll have an epiphany and never want to go back to a point-and-click GUI.

Here's a clue for all the cult-of-linux followers out there. Most people HATE change. Flexibility is spelled c-o-n-f-u-s-i-o-n. Powerful tools are d-a-n-g-e-r-o-u-s. Configuration options are a t-a-r-p-i-t.

Memorize this commandment:
EASE OF USE is ***all*** that matters!

Until conversion to Linux represents LESS change for the average user to deal with than an upgrade to the next version of Windows, 90+ percent of the population won't touch it.

Or, you could go for an even lower common denominator, and develop XXX-windows with built-in pr0n. After all, it was x-rated content that created the market for VCRs and cable TV ;-)

Desktop and Server technique convergence

[CrazedWalrus]

I'm sure it can be done, perhaps by remotely mounting common application and /home folders to a central server. But I've never seen any Howto's or even descriptions of anyone having done this in the enterprise before. Not to say it hasn't been done, just that noone's written how it's done (that I've been able to find).

Why is it that people think Desktop Linux and Server Linux are different animals when it comes to enterprise setups? Enterprises have been doing rapid deployment, diskless (or minimally local), network boot unix installs for ages.

HOW should linux desktops be set up in an enterprise? Exactly the same way as the *ix servers! Any enterprise unix admin worth their salt already has this worked out. The only difference is which applications get installed.

Need that latest patch deployed to all 1.7 bazillion desktops? Update the filesystem that the desktops are booting to and update all of them at once. Messing with symbolic links makes it easy to swing a link back to the old version and reboot the workstations if something goes bad.

Mounting remote filesystems allows users to write their files directly to the network, where it will be backed up according to firm policy. Mounting the system filesystems ensures that every machine is running the latest and greatest. Deploy your apps in OpenAFS, and you can control access to apps via ACL groups.

Re:Desktop and Server technique convergence

[John the Kiwi]

I know this can be done. My question is how?

Where is the information? Where are the success stories with Howto's? What symbolic links should I mess with?

It's all very well to talk about AFS and ACLs and updating a bazillion desktops but you've given me nothing. Got any links to any of this? Bonus points for finding links and information that shows good ways to integrate this with Samba and CIFs to support current Windows based wiorkstations while we integrate Linux based desktops.

So thanks for your post, but unfortunately I have to rate it -1 uninformative.

John the Kiwi

Re:Desktop and Server technique convergence

[CrazedWalrus]

Well, then, maybe I can salvage my KiwiKarma a little...

A good place to start would be Linux Terminal Server Project (click 'Documentation' on the left). Even if it's not exactly what you're looking for, it'll teach you a lot about setting up thin clients, DHCP server, diskless PXE (network) boots via TFTP, mounting root NFS filesystems, etc. They tell you all of this in the context of setting up LTSP, but most of it is general knowledge, and very transferrable to whatever purpose you had in mind.

It might not be everything you need, but I bet it's enough to get you mostly there. At the very least, you'll know what to look for.

As for the symbolic links, that's a little trick I've seen in various enterprise setups, allowing transparent upgrades from the client side, but allowing multiple concurrent installs on the fileserver side.

Example is Firefox. /path/to/apps/firefox/1.07 /path/to/apps/firefox/1.5

Until a couple of days ago, I would also have on the version level a symbolic link called 'prod' or 'current' that pointed to 1.07. Upon installing 1.5 this week, I would test it (maybe via a 'testing' symbolic link, or running from the versioned directory directly). If it works, I would swing the symbolic link 'current' to point to 1.5.

All of the desktops can run firefox via: /path/to/apps/firefox/current/firefox

So the next time they start firefox, they're instantly upgraded -- without reboot.

This can also be applied to network boots. Since you are exporting a path via NFS, you can use the release symbolic link to upgrade whole installations (though, if I recall, there might be some options to tell NFS to follow the symbolic link, since it's a potential security risk. I don't remember, to be honest.) Overnight, once the new system release has been tested, change the 'release link' to the new system version. Things break? Change it back and reboot the workstations.

For large companies, NFS probably will not be the solution you're looking for, but the principles apply no matter what you choose -- AFS, SMB, NFS, etc.

As I noted before, AFS is nice for enterprises, because it a) can be highly redundant, and b) supports filesystem group ACLs. It's also significantly more complicated than NFS/SMB, but allows for lots of neat tricks if you're willing to learn it.

Unfortunately, what you need to read depends on what you need to know, which is dependent on what you want to set up. Fortunately, the system is modular and HOWTO's for popular components are easy to find. Once you get the system running, the rest is just system administration as you would on an individual box.

Hope this helps a little more.

Homer?

[MarkGriz]

duffbeer703 asks: "My employer runs alot of desktop and laptop computers -- something in the neighborhood of 40,000 PCs

Wow, I didn't realize the Springfield Nuclear Power Plant employed so many people.

Forget the OS for now, start with the applications

[SCull]

I think the key is to make the transition process as smooth as possible.

One department at a time is a good strategy, but I'd go even further:

Your users will gonna have to learn to use a lot of new softwares and they won't be happy with that. If all those changes appends all at once, there will be a perception that Linux is hard and complicated. And it will fail.

I'd replace one aplication at a time for as long as I can. Internet explorer would be the first (replaced by firefox), then maybe Outlook (by Thunderbird), and then I'd go with MS-Office (one component at a time, replaced by OpenOffice2).

The idea is to make them learn while they are using there "confortable Windows environment". That way they will be able to go back to there old software (for a while) if something doesn't work as expeceted or if they are in a rush.

And make sure some of the IT staff makes the changes before everyone elses, so they can offer some supports.

Anyway, I which you good luck !

Re:Out of compliance?

[rbochan]

...A mistake was made and a disgruntled employee noticed and reported it to the BSA.

Even nicer was the fact that the same former employee was responsible for keeping the licensing info.

Making Linux an Option

[awkScooby]

1. Make sure you investigate Microsoft licensing issues. In our environment, we would have to purchase a CAL for every Linux, OS X and other *NIX system that wanted to play in the Active Directory. Just because you technically can do something doesn't mean you're legally allowed to do it. Microsoft licensing is extremely complex.

2. Decide on a method for authentication. I suggest using Kerberos 5, since that's what Active Directory uses. You must make a choice -- use Active Directory as your KDC, or use MIT or Heimdal as your KDC with a trust between it and the Active Directory. Due to licensing, and technical reasons, we use an MIT KDC, with a 1 way trust (AD trusts the MIT KDC, the MIT KDC doesn't trust AD). The technical reasons boil down to:

• Microsoft only supports DES-CBC-CRC, DES-CBC-MD5, RC4-HMAC encryption types. Lots of existing Kerberos clients don't support RC4-HMAC, which leaves DES. Yuk.
• Microsoft's Kerberos adds a PAC field to the ticket, which can make for very large kerberos tickets. Lots of existing Kerberos clients have problems with this.

Note that you could choose to have Windows systems authenticate against the AD or authenticate against the MIT Keberos realm, and have non-Windows systems use an MIT KDC.

2. Redirect passwd file lookups to LDAP. You already have an LDAP server -- Active Directory. You'll need to add the LDAP schema defined in RFC 2307, and will need to add the posixAccount auxillary class to all of your users. Part of that process involves putting the passwd file information like uid, gid, geckos, homeDirectory, and shell information in the appropriate attributes.

Again, due to licensing issues, and the fact that we already had an enterprise LDAP directory, we chose to not use Active Directory for this purpose. But, it certainly can be done.

3. On the linux desktop systems, use pam_krb5 to redirect authentication to kerberos, and configure nsswitch.conf and ldap.conf to redirect passwd file lookups to LDAP. On RedHat systems, you can do it all from authconfig, although I think it's helpful to know the files involved.

4. I like pam_access for restricting who is allowed to log in on a given workstation. pam_access can restrict to members of groups, and those groups can be posixGroup objects in LDAP/Active Directory.

I think it's helpful to have home directories on a central server. We use OpenAFS. I don't know if it's possible to have a user's home directory on a Microsoft share or not. If not, you'll probably still be in the business of creating home directories on desktops. Microsoft has some NFS thing for Windows. I haven't used it, so I'll refrain from commenting, other than to remind you to research potential licensing issues.

A lot of this will work across a number of platforms. I have it working on Linux and OS X.

Beyond the stuff above, for managing lots of Linux desktops there are lots of options, but they're probably all roll your own type things. If you have a few standard configurations, you could use rsync. Or have them all point to a central YUM repository, or... Well, there are tons of ways. I can't give you a postmortem on that, because we don't have lots of Linux desktops in our environment yet. Centralized management doesn't make sense for the few that we have.

Summary: pam_krb5 + pam_access + nsswitch + central filesystem == HAPPY

Read up on kerberos. There's a fair amount to get your head around. If you can explain why kerberos authentication is better than "ldap authentication" you should be in pretty good shape.

A Few Successful Hybrid Environments

[Anti-Trend]

In both successful Win/Lin environments I've worked in, one of the key ingredients was the presence of Linux on the servers ...or at least a majority of them. In each case, the Windows domains were handled by Samba PDCs, which also ran NFS and NIS services for the Linux clients, each running from the same consolidated database. That way it's no harder to manage the Linux clients, as you're not asking them to act as Windows clients, but as UNIX ones. If you already have experienced UNIX pros onboard, it shouldn't be a problem for you at all. That way I've found it's much easier to keep things running optimally while still maintaining a single domain mangement base. In fact, one of the two shops I mentioned found Linux worked well enough for them that they are now at ~90% migration to Linux, with only upper management still running non-*nix OSes (seems they're always the hardest to get up to speed, doesn't it?). Of course, I have seen environments where Linux clients are grafted onto existing Windows networks, and they seemed to do quite well. Things are not as seamless as they could be, but everything works for the most part.

-AT

Why _not_ ask Slashdot?

[Anti-Trend]

"One important clue is that it seems like they are using Ask Slashdot as a their first tool to investigate a major business decision."

Pop Quiz: Name a place where you can ask a question of a couple hundred thousand UNIX, Linux, Windows, and other IT/IS pros all in one place, many of whom may have experience relevant to your company's situation. In addition, this place cannot charge a fee for their services, and the answers must be diverse and rapid.

...

I don't know about you, but the only place that comes to mind is Slashdot. Vendors and consulting firms often have a bias to whatever products/services they feel they can offer at greatest advantage to themselves, and not necessarily to your best interest. With all that in mind, I think Slashdot is a pretty good place to start, like bouncing an idea off of a skilled friend except on a much larger scale. That's not to say they shouldn't do their own homework from that point onward, but a company doesn't often get to the point of having 40,000 workstations by sheer guesswork.

-AT

Linux desktop at FTSE 100 company

[weaselprince]

Background: I work (indirectly) for a FTSE100 company in the UK. Last year I architected and deployed approximately 100 Linux desktops to a group of highly technical users, migrating them away from Sun Solaris. I've worked with most flavours of *nix over the last 10 years so regard myself as reasonably experienced systems administrator. Of course that's all relative :)

The technical stuff: Users were running on ageing Sun hardware with relatively low performance (Blade 1000s, Ultra 60s). The applications they run are technical applications for which ports exist for both Solaris and Linux. The new hardware is high-end HP workstations with more memory and processors than you can shake a stick at, combined with Nvidia FX3000/3400/3450 GFX cards. OS is RedHat 3.0. That was forced upon us by the key application which is only supported on that distribution.

Rationale behind the move: Move to Linux because the applications run faster. That's it.

So what worked well?

The major factor in the success of this rollout was the relatively low degree of change in terms of what was presented to the users. The applications they use were simply ports of the Solaris versions. Nothing new to learn. The only difference is that they work a whole bunch faster. Instantly the user base is won over and there's buy-in.

Another, seemingly small, item was the look of the login screen and the desktop environment on first login. First impressions do matter, and getting this right turned out to be very good PR. As the desktops were deployed, users would crowd round the first of the new systems in their areas and "kick its tyres". People were genuinely interested in what they were seeing, and a buzz spread round quickly. On our feedback forms many commented on how much they liked the new, tricked out, environment. In reality little had changed in terms of usability and people weren't frustrated that they couldn't find their favourite application (or analog, where none existed)

There was a relatively low impact for the support team too. Accustomed to Sun's jumpstart, kickstart is an intuitive and easy mechanism for deploying to a large number of identical desktops very easily. Power on, press F12, and the whole thing is automated from that point onwards.

What didn't work well?

The desktop environment was customised from the standard Redhat KDE login so that the right click menu displays a cascaded list of technical applications. Non-essential stuff was removed. Working out how the KDE menuing system hangs together wasted 2 days of my time. Redhat support were useless and I had to use a combination of strace and the source to prove definitively how it works. My major gripe with this whole process was the total lack of adequate documentation. If you're coming from a commercial Unix vendor's platform you'll be accustomed to good quality documentation that gives you all you need to deploy in a couple of hours. Just compare the CDE guides on docs.sun.com to the KDE manuals on www.kde.org and you will see what I mean. This is a fundamental weakness in the OSS world that must improve before large organisations will consider widescale deployments.

What else?

There was no desire or justification to migrate the backend office applications to the Linux desktop. Don't go there - it's a hiding to nothing. If the rest of your enterprise is using MS Office and Exchange there is no sense in trying to fudge things with OpenOffice or Evolution or their ilk. If you do, you *will* have problems. Somethings just don't work, and the support team don't want to spend the rest eternity trying to figure out why a particularly obtuse Word document with some recondite macro is refusing to display in OpenOffice. So how do those users get their standard office tools? Citrix. It just works. Leave the pain of MS support to the masochists and get on with your day job

Re:Just why did Ernie get busted?

[airblaine]

I work for Ernie Ball's IT department and was here 5 years ago when the shit hit the fan. The rockin' on article describes it all very well. The truth is, we downloaded the BSA tools and they DIDN'T WORK. I had proof that I had downloaded them a couple of weeks before the armed guards came storming in. It didn't matter to them. We have been completely M$ free since then and have had no problems at all.

Re:Microsoft writing Slashdot titles?

[16K Ram Pack]

When I've put my systems live, I normally stand in front of a large banner with the words "Mission Accomplished", before spending a few years fixing a broken system that was poorly designed and planned for.

I admin about 30 linux desktops for normal users.

[Victor Tramp]

I'll be suprised if anyone reads this, or even believes it, but..

It's been my job professionally for about 5 years to manage Linux on the desktop for a biomedical company who designs their own ASICs and PCBs on Mentor Graphics..

First I migrated them away from HP-UX as Mentor Graphics ported more and more of their tools to Linux.. It was more cost effective to get brand new Dell machines running redhat [which gets replaced, because Dell's redhat install is crappy] for US$1800 than refurbed old PA-RISC workstations at US$5000+ a box.

Basically, I'm using:

Distro: Debian
GUI Xfree86 or Xorg [depending on which box]
Desktop Env: KDE [muggles love KDE]
Mail: Evolution and MS Outlook [a la crossover office, what a lifesaver!]
Web: Mozilla or Firefox [both are installed]
Office suite: OpenOffice 2.0 and MS Office 2000 [a la crossover office], planner, & MS Project
Music: amaroK or xmms, or whatever they want

The home directories, and proprietary Mentor software are all NFS automounted [it's fine, really], so the only data on the drives is the os and application data.. i lose a drive, no big deal, when the drive's replaced, i reinstall stock debian..

However, I -have- used apt-move [and apt-proxy] to make my own distro of debian in-house for building workstations.. it's a lot more convenient to install the netinst CD and be done, reboot the machine, point the sources.list at the internal repository, load aptitude, and just hit + on the top levels [which essentially loads everything in the repo.], and bam. Installation would probably be more efficient if i used something like Fully Automatic Install (FAI), but i haven't been smart enough to figure out how to get it installed.. It's only good for installs tho.. I maintain the separate repo for upgrades.. That and it keeps people from installing things willy-nilly from the net if the only repos the workstations have is an internal server with a subset.

User accounts are all managed via LDAP, tied in with the corporate ADS directory, one login to auth them all!

box configuration management is all handled by Cfengine2 -- all hail the university of oslo! Cfengine has made it possible for me to manage all the boxes at once, no matter what the hardware discrepancies.. i can do the work of 4 people, by myself. that and, the configuration repairs itself if discrepancies show up on the workstations for some reason..

Let's see, that's software centralization, user data centralization, user auth centalization, and workstation configuration centralization.. for 30+ boxes across two buildings, for engineers and managers, for almost 5 years.. There may be better ways to design a network of workstations, but the support model I've implemented has really worked out for me..

Once a "normal" user gets used to the idea that the computer will do whatever they want it to do, even though it's not windows, the questions about how to do things taper off after a while as they get used to the new system..

It's sad really. I've been doing the Linux desktop professionally longer than anyone i've ever even heard of. I know for a fact the Linux desktop is completely viable, but nobody seems to believe it. Non-technical and technical people alike all seem to have their doubts and never get up the energy to actually explore it.. Heh..

anyway, good luck