The problem here is that you're talking about the difference between needing to respect constraints and being able to work outside of those constraints to set them in the first place.

If you take the simple Windows example of building a local service, you can see the complications for the non-admin developer. First off, how do you install the service that you've just built? The service can run in a local context, but it needs to be installed using an admin context. Now if I can't install the service, how do I test it? Can I attach a debugger, do I have those permissions?

Note that this isn't an issue of getting the service onto the user's computer, the sys admin can manage that process. This is a matter of being able to pretend to be an admin to get stuff working on my own computer before I push it out to users. Now it is possible to manage permissions in such a way that I can actually test what I'm doing without needing full admin permissions. But in most domains the simplest way to do this is to simply provide admin permissions (to the machine, not the domain) and let the devs support themselves.

The other major option is to set up some VMs for doing this, but then you have the problem of needing to create network test users so that the dev can manage the machine but still access resources on the network as if they were local. So it's a similar problem in a different place.

Yes we have several installs at my office.

We run hundreds of Linux servers on EC2 and much of the original software was written in .NET. Most of our commodity servers have therefore been ported to Mono. Several of our core servers are running Windows. So we end up using Mono when we need scale (to save on licenses) and .NET on Windows for ease of implementation (like Services) on those servers that won't need to scale heavily.

At the end of the day our team uses both VS and MonoDevelop and I don't really think that either of them are going away :)

I'll add to this comment, FTA:

Back-end services that require the performance are implemente in C++.

And most of these 30k servers may be "running" PHP, but goodness their code is not all PHP. Facebook leverages a large amount of open source software and it's definitely not all PHP. MySQL isn't PHP, MemcacheD isn't PHP, Cassandra & Hive are written in Java, Thrift and Scribe are built in C++.

This guy's whole premise is completely under-researched, to tell people at Facebook that they're missing a 10:1 performance opportunity b/c of PHP is pretty ridiculous.

And that's the problem right there. The primary purpose of the IDE is to leverage computer power to improve developer productivity. If you're not using an IDE with re-factoring tools and test suites and code generation and maybe some handy UIs... then you don't need that much RAM.

Of course, it could also be argued that you're wasting a lot of money. RAM is cheap, your time isn't supposed to be.

Based on the original post, I think something went wrong during the talk with the boss: "I asked my current employer and was shot down". The question here isn't "can I?", but "what can I do to?". Unless you walked into that room with ways to make it happen, then you were dooming yourself. Go read 4-Hour Work Week for some inspiration / ideas.

As a guy who spends 20+ hours / week on Facebook (it's my job (no I'm not kidding))... Facebook and MySpace are very different.

My friends and family are mostly regular Facebook users (yes that includes 50-year old aunts and uncles). Facebook has a 35-40% penetration in Canada, they didn't get there without having a few people over 35 :)