Yes, it is possible, while you usually do not want it. It is not easy to get a server system scalable for a whole school, with service contracts, warranty and stability for this application. The central application server requires hardware that fits pretty much into what Sun offers with their machines and solutions. If you are in a university, going with Linux here would be cool, as you would have lots of students to hire for working on this. But in a school, you do not want to have more trouble then saying "please get us 25% more RAM into our server and the latest software update" to the school secretary, so that she calls and gets the service installing this in less then a work-day...

The size of your Photoshop file should not matter, as it stays on the server. And you will have a server with a lot of memory. By the way, your use case is a great example - have 100 clients, and 40 of them running Photoshop, the others doing word processing and such. In a typical fat client set-up, you'd have to equip 100clients with a GB of memory, so that they are all suitable for all tasks (you do not want to have different configuration in a school, just one model for students, cause you will need spare systems, installations, servicing and purchasing procedures for each model you get). The server set-up just needs enough memory for the simultanously running apps. In other words, 100 fat clients here mean 100GB RAM, thin clients + server less then half of it.

Hi, first, do not make any calculation based on buying machines, but on maintaining them. It maybe easy for you to get a budget for 100 laptops. But if these laptops are carried around by the kids, are to be used with current software, to be replaced every two years (max. life-span I would give to a mobile device under such conditions), you will need at least three additional stadd for doing just that. And how easy is it to get the budget for hiring staff? You do not want to have teachers, who should care about the kids, use their time in doing it administration and servicing, right? I think my proposal would be a real thin client. A perfect solution may be something like Sun's Ray, as there is NO (zero) software installed on the client. This means that replacing a broken terminal requires you to plug-in one, that's it. Can be done by everyone who is able to change a light-bulb. Sun has huge reductions for educational institutions by the way. Attach this to a relieable server with some failover and such. Spend 2/3 of your it-budget in this, including its maintainance. Best is to get it with a service contract for the expected life-span. Do not even think about building this on your own or saving some dollars here, again, you do not want your teachers spend theit time on computers later. While the work-hour of someone servicing a centralized server is usually higher then the salary of a guy offering you to take care of laptops for some dollars, there will be only one server to maintain. Every software installation will be done only once. You will not have kids with laptop bags waiting in queues to get their software updated. You can scale it up as needed (at least if you take a machine that is designed for this and not some PC in a server box). Add memory, CPU and such to the server when you need it, install the latest software, make backups - but NEVER have the it service and the kids in the same room ;-) The only reason for laptops is if kids really have to use it at home. But again, such a machine will be dead after two years, and replacement parts and servicing costs are high, so this will eat up your it and your human ressources budget.

Ok, we are back at the point where people who do not see a feature they need in blender start to invent a new app, often a rewrite a commercial one that we do not want to pay for. Here starts the trouble, so what is a parametric modeler? By definition it is creating models from parameters, thus changing parameters reflected in the model. You can do all kind of stuff with that idea, in 2d, 3d, on meshes, solids, generic objects, text... it does not define a cad yet. I see two successful concepts of 3d applications. One is offering a great toolbox for one field of modeling, e.g. a great solid modeler, a cool 2d sketching CAD, a powerful mesh modeler. The user than uses tools from a pallette of apps, modeling solids in the modeler, drawing a plan in 2d cad etc, finally importing everything for rendering and output (rendering can be a raytracer or a 2d drafting/layout tool). The other way is not to write apps for the type of data that is worked on (solid, surface mesh etc) but the useage expected. As I understand that is how Blender developed, a tool for 3d animations. With that approach, the tool gets almost unuseable for anyone else of course, and here is the reason for most complaints about render. Do you want to make design (scale of mm, m, km?) requiring accuracy, simulation requiring simple but clean solid models, animation with high polygon count surface models and anim support, game design with low polygon count...? Too many try to put everything for everyone's useage scenario into an application, that won't work. I think in open source the tool pallette concept works better, we have ways to do solid/csg stuff, surface meshes etc. The missing parts would be the solid cad package (that is what pro/engineer would refer to) and a really good 2d cad able to exchange data with the 3d foss tools.