If you spend more than $15 on a thin client device then you are a fool!

The push toward more expensive thin client devices is getting out of hand. There's no reason they need to cost that much.


I am appalled and disgusted by how expensive thin client devices are and how much they effort they are to manage. Unless you specifically need to watch videos on these things (which I guess applies to about 3% of the people who buy them), I think anyone who spends more than $15 on a thin client device is an absolute fool.

I have an alternate suggestion:

  1. Go to eBay and buy the cheapest used thin clients you can. (5 for $20, 3 for $13, etc.)
  2. As they come in the door, configure them to use DHCP and have them automatically connect to a published desktop called “desktop” using a DNS name server location called “citrix.”

Users turn these things on and they immediately connect to a published desktop on a Citrix Presentation Server. Upon logging on, Program Neighborhood Agent software on the server drops icons for various application on the remote desktop. ICA Passthrough (which works very well and does NOT negatively affect performance) can be used to make a double-hop connection to the applications on its own silos. (Or if you don’t have a siloed environment then simply have the thin clients run published desktops on member Citrix servers.)

That’s all there is to it. No management. No client viruses. No additional firmware management software to learn. No client-side settings that will ever need to be reconfigured. While this idea sounds crazy at first, I challenge you to find something wrong with it? Seriously... what’s wrong with this idea?

New thin client devices from the big vendors are expensive. The “manageable” ones just introduce another management application into the lives of overworked admins. This is a total sham. These new “features” the thin client vendors talk about are mainly ways to keep themselves in business. (XP Embedded? Local Media Player? 256MB of RAM? Do we really need this to run Outlook?)

Any late 1990s-era Windows-based terminal that runs Windows CE 2.12 will be fine in today’s environment. You can pay a high school kid a few dollars an hour to receive the new devices and use Google to figure out how to configure the firmware for DHCP and the default auto connection.

So seriously, what’s wrong with this? Some people have suggested:

“But you can’t manage these devices”

So what? What is there to manage? They’re freaking thin client devices. Who says we have to manage them? Do I need to update firmware? No. I just need them to run ICA. They use DHCP and locate their servers based on DNS. That will never need to change which means you don’t need to manage them. The only reason you need to manage new thin clients is because they’re bloated and break too much or weren't configured properly in the first place.

“But you can’t put the newest ICA software on them?”

So what? Why do I need ICA 8 or 9? For the past 7 years ICA has supported 24-bit color and plenty of resolution. What else do you need?

“But you can only run one session”

So what? I’m connecting to a desktop tier (which I can manage and is the level at which I want to manage my environment).

“But this is just wrong!”

No, paying $500 for a thin client device that requires management software and can get infected by worms is wrong.

“But they won’t have the same performance as new thin client devices.”

Performance for what? Word? Outlook? Guess what? They were good enough for ICA sessions in 1998 and they’re still good enough today. Even though the applications have become bigger, pixels are still pixels as far as the thin clients are concerned.

“But these things are old, used, and have no warranty.”

Yeah, there are also unlimited supplies of them and they cost $6 shipped. Buy extras.

“But my rep from <thin client vendor X> won’t invite me to the golf outing next year.”

Take the money you save on thin client hardware and the time you spend on management and buy everyone in your IT department a new set of clubs and a country club membership.

Seriously, look on eBay right now. (Dig deep. Sort by price.) If you ever spend more than $15 for a thin client device then you’re a fool.

Don’t be bamboozled!


Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

I have never seen the draw of these. I have had many clients that when you try to suggest they spend $500 on a "thin client device", they respond by stating they can get full blown PCs for $350 from Dell.
You're absolutely right, it's crazy to pay 500$ for a thin client.
As we're playing with thin clients (95% of our customers) for more than 10 years, let me just give our feedback
Why companies use thin clients ?

1. Meantime between failure of a thin client is +/- 1.000.000 hours
-> for a PC, it's +/- 100.000 hours (10 times less !)
2. No maintenance needed on thin client, no viruses,... (I mean "real thin clients", not XP embedded ones...)
3. Electrical consumption (minimum 2 times less than a PC)
4. User can't break it (no floppy, no cdrom,....)
5. thin client are less stolen than pc's as they can't run alone (some customers have up to 20 pc's stolen / month !!!!!!)

So why having a new thin client with a lot of features that cost 500 $

I would say : NO REASON as 95% of new features are not needed BUT....
I do not agree when you say you can always use on old thin client....

Having a thin client with a terminal emulator (3270 or 5250) built in can be very usefull in case of migration from a mainframe terminals to SBC
Having a LPD included is very usefull as you can connect printers on LPT port or even on USB port and the thin client will act as an LPD or be able to use TCP/IP printing (as HP jetdirect -> cost saving...)
It's much faster to print via lpd or tcp/ip printing than capturing the local printer...
New thin clients have a VIA CPU of 800 Mhz or even 1Gz and combined with latest client + local internet explorer or flash player.....multmedia experience is much better (and more and more needed...)
For me, the management part should not be considered when buying as there's a great chance that you will never have to manage them remotely....

and don't forget that thin clients must be able to support other things than just a full desktop connection
Often, you must connect ticket printers on COM ports (not all works the same way), or have a smart card reader connected,.....

BUT, a very good thin client with all these necessary features cost approx 200 $
Don't spend more on a thin client....

If you want, we could do a market review on all thin clients together as we nearly know all of them

I been doing this for years. For basic desktop apps, this solution rules.
I'm still using compaq T1010's from the last decade with no issues....
Well I could give you a few reasons why you need a managed thin client. The first idiot who walks in off the street and learns how to reset the terminal by pushing ctrl g or whatever the sequence is and you have to go out and drive 50 miles to reconfigure a unit. Don't think it won't happen a few hundred times. Where is the savings there?

If you run a large network with DHCP, you haven't a clue where the bozo who is sending nasties to the CEO is sitting. Especially if you use anonymous logons. Any network manager worth a pound of salt wouldn't let dhcp run on his network openly because he knows he is asking for every idiot to come in off the street and plug in their laptop and be on their network and bring in outside viruses.

When these old things blow up and start on fire, you can explain to your CEO why you spent $15 on a device that just burned down your building. I've seen at least 20 old wyse units do this by the way...luckily a fire extinguisher was handy.

Sometimes going the cheapest way out is not always the smartest.
Deploying applications, yes. Deploying desktops, no.
Deploying applications, yes. Deploying desktops, no.
Deploying applications, yes. Deploying desktops, no.
I use Sun Rays from ebay (AU $80ish each) including a 15" LCD
you need either a Linux / Solaris server (ok another point of management) but you also get 9.0 series ICA clients (linux / 8.2 on Solaris)
I actually use Sun Java Desktop on them for everything except when I need to run windows apps)
I agree, Thin client devices are too expensive.

TSCALs are also too expensive.

There is only so much pie to go around, and MS gets the biggest peice first.
Dude, you need to spend sometime in healthcare... We have doctors and nurses who smooth-roam... Try doing that with an auto-connections. How do you get sent to right server when the Citrix doesn't know who you are yet? WI works great on thin-clients, but not old ones. Users expect certain keys to work with in their session and not for the device, ICA 9 fixes alot of these issues.
No, Smoothroam works fine with this method. The cheap thin client autoconnects to a random server. The user logs on with their real credentials to that server, and PN Agent running on that server connects the user back to their Smoothroam / workspace control sessions. I actually have two customers doing this exact thing right now.
Why not deploying desktops ?
Overall I agree with you Brian. I've been on a crusade the last few years to get customers to think about what they really want and find a reseller who can provide that. So in terms of Thin Clients: press a BIG button to power on, enter that ol' password and you should be done. This will do for usually 70% of your users. Then there's users that NEED to have scanners, smart card readers etc. Buy full featured one's for that (NOT meaning a trimmed down version of Windows).
This should work if your users arent to picky about performance (If jpg's start appearing in 12 segements over 2 seconds then most users start nagging).

Someone already mentioned that it's almost cheapers to buy a fully-featured PC from Dell than it is to buy a new Thin Client with Win XP Embedded w/ SP2. What's wrong with that picture huh? ;-)

Lastly, SBC in the end is all about saving money (TCO or whatever you like). If you want to do it right, then do it like it's ment to be done. Deploy applications server based and buy DUMB terminals. Don't do both, you'll be spending twice the money!
Really great points! I totally agree.

1. I think the Eiger / Monch stuff from MS will really give people a great option vs. traditional thin clients.

2. I really agree on the "if you're using SBC, then use SBC!" Don't spend the money on both ends.
It amazes me how silly this whole debate is. Surely if a user requires local applications then they need a PC. If not, then they can use a Thin Client. There is absolutley no point giving a user who only requires basic bog standard applications a "Fat" Thin Client - give them the cheapest bit of tin and spend the cash on the back end. If they need local processing, get them a PC with an ICA client on it and be done with it. Anyone who wants to argue that a terminal with a local OS that runs other apps is not effectivly a PC anyway needs to wake up and think about the whole point of a Thin Client anyway in my opinion.
I think this is the point I was trying to make actually. I like the "cheapest piece of tin" approach, and that's what I'm advocating. (Or maybe you were agreeing with the article and responding to some of the other comments?)
I was of course responding to the comments Brian, and agree with what you are advocating entirely.
"The first idiot who walks in off the street and learns how to reset the terminal by pushing ctrl g or whatever the sequence is and you have to go out and drive 50 miles to reconfigure a unit" If you are buying used you can afford to have spares onsite and ready to go or you can walk the user through the configuration over the phone.

"Any network manager worth a pound of salt wouldn't let DHCP run on his network openly because he knows he is asking for every idiot to come in off the street and plug in their laptop and be on their network and bring in outside viruses." WTF!

We have over 3,000 workstations on our network and we shouldn't run DHCP? There are other ways to prevent unknowns from attaching to your network.

I like Brian's idea and we are going to test it out. Brian thanks for getting "fired up" and causing this lively discussion.
How many cards will Brian get this year from the thin client manafacturers?


You vote :)
And session reliability?

As far as smooth-roaming and PNAgent... So how did you fix the issue with PNAgent trying to smooth-roam the previous desktop?
So, I should go justify a FAT client because the department needs Media Player for one application that users will use once a year.
So it's all or nothing? Not every thing runs well under SBC, but the majority of the things can. So should I scrap SBC across the enterprise for that one poorly written app (sorry softgrid is not the answer here) that will eventually be replaced?
What about session-reliability? Mobile users that use carts and dr. that use tablets/thin laptops?
Overall I agree with you, I think electronic company’s just missing in design. It's so simple just make few systems files and antivirus systems located in RAM, it will prevent device from any infection or system fail. Additionally, it would be simple to include one more chip which will provide a remote connection to this device which give possibility configured it from anywhere.
Even there can make a device like ThinPoint where the terminal will be connected and this device can verify the system, configuration, connection i.e. and change it automatically so it’s just thin robot. Idea is not soft-soft-soft, is chip-chip-chip much better. Txnks
Has anyone ever put a hard Cost of ownership figure to a Thin client vs. a PC, I have been trying to put some realistic numbers together for management. All things considered (Software cost, hardware costs, maintnance, personel), I figure it cost my company (being conservative, assuming the bargin basement $399 dell PC) PC: $700 Thin Client: $350 (We exclusively use HP t3310s). Then the ongoing maintnance PC: $200/month if there are no problems. Thin Client: $50/month, with or without problems. I still havent spent any mony on monitors. I would like to see what everything thinks of these numbers, granted we dont buy a $399 PC, Just for sake of arguement. Also I would consider the cost of maintnance on Laptops to be signifcatly higher, esp if they connect to other networks.
Now this is an interesting article and series of comments that I would like to put in my two cents. I agree with that fact that the PC is going to cost more money and a lot, in the long run, and not only money but time. I understand and somewhat agree with most of the other comments but the true facts are that PCs COST MORE MONEY. They really do.

For example, now days I run a small company and we are distributed across the US. My operations manager is having major troubles. He is not productive and he is mad. I need for him to send me his computer and then I have to take MY time to fix it and send it back, all the while he is not productive. Now, if I had a thin-client it would be a hell of a lot easier and the time and money savings are truly huge. This is an example of a problem that I have right now and if it happens for me a really small company then you know it happens in larger environments, no matter how well they are run.

Heck, I’ve been doing this a long time too and I have many customers that run with thin-clients and if they go about not just buying the cheapest but buying a thin-client that is designed and configured for their environment then they are almost always happy and save huge dollars.

The last thing I think of is that when we, as engineers, think of solutions we tend to think, “How can I solve the problem I have” and not, “how can I solve the root cause of the problem”. We, as engineers, say we can fix it with this piece of software which in most cases I see as placing a band-aid on top of the problem. Fixing something by slapping something on top = band-aid. I just see thin-clients as a root cause fix for people who do not need a PC and my operations manager does NOT need a PC to be productive and I bet if you look at the end-users you have you will notice that a huge amount of them do not either. Thus a thin-client is something very useful, if designed and configured properly. Of course, it can be argued that a PC, if configured and designed properly, can provide great value too but then you get back in to cost.

Anyhoo, I do like thin-clients. I really do and always have. I use a thin-client laptop a lot.

Just my two cents… I do however understand and appreciate yours.
Our users need to have a FAT PC so that they can use Media Player to allow them to see a Christmas message from the CEO every year. Which half of the time does not work anyway because the PC’s get all misconfigured because of users tinkering with the settings. For the rest of the year Media Player is used by the users to play funny videos that their buddies send to them

If only there were a Thin Client that could run and only an ICA/RDP session and have media player so that we could facilitate this.
Overall I agree with Brian on this as well, but within my own compay we need the ability to view streaming media for computer based training that everyone needs to take once a month. Unfortunatly cheap Wyse terminals we've had for years can't perform this function and we are having to look at clients that can support windows media player locally. As a company we refuse to deal with Dell (the CIO had a bad experience with them and wrote them off - oh well) and have entered into an agreement with HP. There is also a negative reaction from people when you tell them they are getting a thin client instead of a computer causing a political fight within the company. You and I both know that it's not a slap in the face to use a thin client, but end users think they are being punished by the company if they aren't getting a computer.

One solution then is that we try to buy inexpensive winterms with no drives that look like a computer and then tell our end users that they have a new high-tech computer. It's stupid, I agree, but in the end the users have device from which they can get their training and the IT department has a way to manage the rest of the applications they are required to have. Just my opinion..
Brian brings up an interesting idea about ebay which we might implement for select customers.

Thin clients have definitely saved me time and my customers money. I have hybrid environments with pc's and thin clients and I rarely if at all have any issues with the thin clients. I have been slowly converting my clients on the idea of thin clients. In many cases a truly thin client is not a reality because of various needs, but hybrid environments work extremely well.

It's funny when you meat a tech who doesn't think thin clients have any worth at all.
I agree Thin clients are to expensive.
As an alternative you can use your old PC's, get rid of the HD and make a PXE boot image (realy like thinstation.sourceforge.net). Now you can centraly manage it and use your old hardware. Make an boot image to start the desktop and users think they work on a pc again :)
Even if you could buy 1000 thin clients from Ebay, most likely they will be a mixed bag of model types with each having their own way of setting them up.

And let's say you figured out how to configure each and every one of them...are you seriously going to consider manually updating every single one??

Instead imagine shipping the thin clients to each remote office and having the locals simply plug the cables in (not rocket science stuff) and when the units appear on the network, the <gasp> management software reconfigures every thin client as they boot up.
Finally someone who gets it. If I want to manage NeoWare devices with their management console, I can't manage other devices as well (which you will get on ebay). Go with a good company (i.e. MaxSpeed, NeoWare) invest the money up front, and the cost savings and the hassles will all be simplified from the second you implement them.
Technically, Brian is correct. We can always use the 80/20 rule with computing, or anything. So, we may end up with a hybrid environment w/ PCs and cheap TCs. Great. Again, works technically. BUT, no VP of IT or CIO can say to his/her board of directors or shareholders, "Look at how great we are, we saved $50K by buying used computing devices from eBay. Oh, are they supported by any manufacturer? Nope. How much did time did we spend to check the USED DEVICES THAT WE DON'T KNOW WHERE THEY CAME FROM to see if there was any malicious code on them? Huh? Um.......but we saved $50K....grin."

Yes, I know what are the chances of malicious code being written for a CE or Linux device? Pretty small. But shouldn't one check? Would you just plug something in your network that came off of eBay? Probably people do it everyday, but is it really good policy?

Just two cents from outside the techie world.
What do you mean "manually updating each one?" What is there to update? This is the whole point of my article.. I don't think you will ever need to update these things.
Brian, "manual update" i.e. - setup RDP/ICA connection info (enable drive redirection, audio, etc.), video resolution, color depth, keyboard mappings, printer redirection settings, etc.

You can't honestly expect to send out thousands of mixed types of thin clients with unknown configured states and expect users to just figure out what what settings they ought to use?
I've done this with thousands before and it's worked fine. All that stuff can be configured via policies on your Citrix servers. All you have to do is set up the thin clients to use DHCP and auto connect to a published application called "desktop" with the server location "citrix" specified and you should never have to touch them again.

Again, this won't work in all cases, but I have two clients with almost 5k thin client devices between the two of them, and they have a totally mixed bag of all different types of clients, and it hasn't negatively affected them at all.

You missed my point. The fact that you spend money to have someone touch the unit at least once (that's 5k touches in your example!) is a good enough reason for most to seriously consider a good remote management solution. It's not how how few clicks you have to make on the thin client's GUI, but the fact you have to have a trained person to go and touch it once.

Also, are you guaranteeing that your example customer of 5k units will *never* have to touch their thin clients again? Touching them a 2nd time can be much more costly than the 1st since the physical units are deployed out in the field, so you have to send technicians out, whereas the 1st touch may have been in a central distribution location.
Ah.. I see.. one touch vs. zero touches..

This is a really interesting conversation. In the cases where I've done this we've used an intern (making $7-10 per hour) to do the first touch. There really isn't too much to know an you can do a unit in about 4 minutes.

Of course no one can guarantee that you won't need a second touch, although I honestly can't think of a reason why you would need to. This model lets you change your servers, change your network, etc, etc... all without impacting the clients that have been deployed.

The bottom line to all who may read this:

You need to do what works for you. You've seen what the vendors recommend. You've seen some of the issues with managing thousands of thin clients and providing client-based access to multimedia content. But, if you don't have too many clients, read the pros and cons in this thread and then make a choice for what works best for you.

Through this whole discussion I think we have failed to touch on the issue of security. I actually LOVE XPe-based thin clients for the PC-like behaviour in many aspects (easy drivers, easy ICA Client upgrades, easy tweaking, etc.) although I will admit it has its downsides (viruses, $$$, etc.)
But at least for me, there is a HUGE value in the whole concept of publishing apps rather than desktops, specially for the security-minded. I will accept, of course, that this could be mitigated using the double-hop ICA Passthrough design, which I also like a lot....
To sum it up:
1) WinCE: Great for single-task (or single-app) users, double-hop deployments, or TS-only environments
2) XPe: Better option for multi-task, Citrix Presentation Server Environments, and multimedia needs.
3) Linux: Check with your customer. If their end users are able to cope with the GUI differences, go for it. (In my experience, customers tend to favor the Windows GUI...)

Now, about dropping management tools altogether: I agree with Brian that you shouldn't need to configure/upgrade the thin clients... but what happens IF you do? Can you be 100% sure that in the next 7-10 years (which is the time I expect the TC to be working) I will NEVER EVER need to touch the TCs configuration or upgrade its firmware?

I wouldn't bet on that, but that is just my opinion...

We can agree to disagree :-)
I would argue if you aren't treating your published apps like publsihed desktops, your not secure at all.
I have clients with plenty of thin client devices with firmware from 1998. (CE 2.12) Seriously, what would you ever need to change on them?
Will this solution work from locations that use just a dsl, cable, or dialup connection from outside of your network? Like a community based home that uses a thin client for accessing a single application. Simplicity is key, they only need to open and print from this application (no smartcards, scanners, ect..). Our locations are spread up to 12 hours away from our central office so you can imagine that the idea of this solution is worth the dumb question.
As an addon to the above question regarding geographically deployed cheap thin clients has anyone had experience with a Xincom twin WAN router? Ordered a test unit but supposedly does inbound/outbound load balancing and failover for anything from DSL to DS3. Thinking of deploying this as a branch office connectivity solution for a customer with about 200 different sites all over the country. Kinda gets back to that "Are we really going to touch it once..." question.
Mike B.
Sun's Sun Ray Ultra Thin Client, managed centrally by a Solaris or Linux server, fits Brian's scenario perfectly. Well, OK, they're a tad more tha $15 but aside from that, about 90% of the issues raised are addressed by the software. I'll take just one as an example: "upgrading the desktop". As Brian points out, there should be no need to do this. HOWEVER, at times a supplier may add useful new features to desktop firmware as Sun did with the recent 3.1 release of the Sun Ray software. With Sun Ray, you install the new firmware on the servers and the desktops pick it up automatically the next time they re-negotiate the DHCP lease. How simple is that? The user is interrupted for less time than a coffee break, the server side restarts whatever Windows connectivity software you are running, Citirix client for example, and you get you Windows desktop back with the insertion point sitting at the last character you typed!

You really need to take a SERIOUS look at Sun Ray.

RE: Use a PC with Linux

Posted by habeekm on 06 June 2005
I agree Thin clients are to expensive.
As an alternative you can use your old PC's, get rid of the HD and make a PXE boot image (realy like thinstation.sourceforge.net). Now you can centraly manage it and use your old hardware. Make an boot image to start the desktop and users think they work on a pc again :)

PXE makes sense. a year ago we started converting our network of 50 old desktops (mixed with processors ranging from pentium-200, celeron 266 to p3-866Mhz) because we have to implement a new custom database software that can only run from at least a windows xp o.s. we bought this half-length pci go_diskless cards from www.x-novation.com. its configured to take control upon detect from bootup so even w the cmos battery is dead it does nothing to the system. the rest i just followed what thinstation.sourceforge.net had instructed. had a window2003 server with terminal server, configured the built-in dhcp in the tftp utility. i had 2 partitions in the server. the primary containing windows2003 which i data protected using a recovery card bought from the same company so that in case of corruption, viruses, etc all i do is reset the server and everything in my primary partition is back to normal. this mean users can still customize their desktops and will retain them as long as the server is on. but once the server is reset everything goes back to the last backup state.  the second partition contains all the user data.
no manual (or automatic)updates needed n clients becoz it will stil run normally even wen the battery dies, i only needed 32 mb ram (the rest of the ram and drives i took out for future repair, after all im using old units and parts are scarce), cool operation (the cpu heats very way low than when it was still fat), unbelievably very fast even when im only using 32mb ram and pentium-200 processor. the updates to windows and the anti-virus are automatically,but temporarily, loaded from the internet. but every few months when patching our database program i have to disable the data recovery program to make the update permanents. this is also the time the windows and anti-virus updates become permanent to the primary drive.
its perfect (only for the company). it lessened IT and maintenance work by more than half. in fact there used to be the 2 of us doing the work.  the other guy was fired at the start of the year.  now im worried the company might fire me too becoz im not functioning as what im originally hired to do other than dusting off the monitors and removing the jammed staple wires in the keyboards once in a while.
Thin Clients are not perfect, but there are some really good reasons to consider them it user's requirements don't dictate the requirement of a full PC, and some general comments:
1. Lower power consumption.  A running PC doing nothing can easilly use 100W+ of power and can use 300W+ when in use, whereas a thin client doing nothing uses a few watts and maxes out at 45W.  Electricity costs money.
2.  Less IT Staff required to support thin clients.  A locked down thin client w/ no moving parts doesn't need a IT person at the location to support it.  Ship the end user a pre-configured thin-cleint that they can plug into their network and turn on.  If they have a problem you can remotely manage the device with VNC or send them a replacement.  I have a lot of Wyse thin-clients that I've never touched after they were deployed.
3.  I've never had an end user do a G Key Reset on a Wyse Thin-client.  Yes it can be done, but users could also erase the disk volumes on a PC w/o much difficulty, even if you password protect the BIOS.
4.  No moving parts in a thin-client increases the MTBF way longer than the 3-5 years you get with a PC before the hard disk, power supply and CPU fans die.
5.  Lack of Video RAM can severely degrade perceived performance, and absolutely mimits the display resolution and color depth.  The old $10-50 thin-clients on EBay often have 1-2MB Video RAM and support 1024x768 @256 colors and some support 16 bit color, but are so slow to paint the screen that you're better with an abacus.   If you require high resolution display, i.e. 1280x1024 on a thin client, you need a newer device with a better graphics card.
6.  If you need to support local apps, it does NOT mean that users need a PC.  I use XPe thin clients with some local applications w/o any issues.  It just depends how large the application is.  Local scanning software is a perfect example of an application that runs fine on an XPe thin-client.
There are users that need a PC, but there are usually an equal or greater number of users that get by just fine w/o a PC or an IT person to support their computer.  As always, choose the best solution to the problem at hand, and don't try to push people to use a computer that doesn't meet their requirements. 
Assuming you have users who use Web Interface most of the time to launch published applications on fat clients, but you want to put some thin clients into the client in addition to what they have so you publish a desktop for each user and set the thin clients to connect to this desktop automatically. The desktop is locked down so once they are connected you want to run PNAgent on the server to give the user a list of applications he/she can launch. How in the startup script (Kix or otherwise) do you launch PNAgent on the server but ONLY when the user connects to the published desktop. You don't want PNAgent started when they launch publised applications on a fat client?
PLEASE PLEASE PLEASE take a moment to tell me how you would do this!
Mike B.
He is right on all accounts.  I've been saying that for along time, but they just won't listen

I did eventually find the way... Just use your logon script to copy a startup shortcut into the users Startup directory. This directory is not processed when using seamless windows (IE the web interface) but is processed when you fire up a desktop...




What about when running windows muiltipoint? S

Totally agree
But they are really slow.
I bought Dell Wyse X150SE. Everything was OK but drawing pictures was really slow (about 0.8 sec). Imagine what a torture was rapid switching between a lot of pictures, while this worked fine when i connected to remote desktop using another computer.
This is not usually an issue with your thin client, or the entire VDI concept. Many VDI setups lack GPU servers, so everything is processed using the CPU and that's a load of crap.

We have clients using GPU intensive servers and remotely they have similar clients and they're doing CAD work etc all day long.