unhappyms

I think they would stay at Google (in its current state). Google is getting big but is still not as big as Microsoft (political). Google has more data (I would imagine that is something search folks want more of). Google seems to be the best environment for top search engineers for a variety of reason. My guess is that the top search folks (along with others in the top of their field) get to choose where they work and what they get paid because they are at the top of their game. So if Microsoft is where they wanted to be they would already be here. If it is where they want to be and they are not shame on us for not recognizing that.

DevDiv makes about $1 billion for MS. Office on the other hand makes about $18 billion. So in terms of politics Office gets to decide what tools they use for the most part. Office sees it as a risk in terms of profit and hitting ship dates on time to switch to the latest stuff DevDiv is spitting out.

Rumor has it WPF is going the way of Windows Forms. It will be supported but no new features. With that in mind, think of all the UI frameworks MS has had wrapping up Win32: MFC, ATL/WTL, WinForms, WPF. Nothing seems to stick. Office just writes their own stuff that they KNOW they can depend on it being there. Duplication sure. But DevDiv is notorious for releasing things and deprecating the past "great" technology within a few years.

There is another reason they ignore the new tools: managed is slow for the client. For the server world managed is great and has won. Nobody in their right mind would write native web apps. They write the stuff that needs to be native (db, webserver) then glue the app together with python, ruby, C#, or Java (I know, a vast oversimplification of webapps).

Clients don't have the advantage of the webapps. We don't have a lot of computer to do things in parallel. Every millisecond counts to make the app appear as responsive as possible.

Why does managed suck for the client? JITing is slow and the code it generates isn't nearly as good as a compiler. NGENing is possible but still slow because of garbage collection and marshaling data between native and managed code. Little things like that add up and slow down responsiveness a lot on the UI side of things and that is what the user notices and complains about.

What I meant by unsustainable is that I feel the retaining talent problem is less related to money. My guess is smart people are leaving mostly because of the environment and not because of pay and benefits. When those smart people leave, theoretically the quality of the product suffers and overtime the competitor with better people wins. If MS profits go down over time, and investing in things like Azure and Bing and Entertainment (all which loose money), I don't see how they can continue to throw money at the talent problem. But man if I could get paid more AND work in a better environment (one where I had an impact) I would go for it in a heart beat. So in that sense money DOES matter. I agree with the rest of what you said though about it being difficult to measure.

This is what I think of when the subject is talented people leaving MS: http://www.techflash.com/seattle/2008/12/Key_data_center_arc...

http://googletesting.blogspot.com/2009/06/james-whittaker-jo...

I think Amazon and Google are better because of it.

I'll admit some of the disdain is related to the process with no consideration for the the actual product itself. The taste in my mouth after Dev10 makes me unhappy and my desire to leave MS seems to align with yours: "all over again for Dev11." Good luck in your future endeavorer.

Google/MS/Amazon do have some VERY smart developers. But there are dolts too at the big ones. The fact that you realize there are better developers than you puts you above most. A lot at MS think that because they work at MS they are the best thing in the world for software.

Give larger raises with promotions would be my guess. MS likes to promote (expect more) and then give below inflation raises.

Very true. A lot of 2010 perf hit is because of WPF. Real hardware and a good video card is required. I never dreamed that a decent video card would be required for a text editor but it is.

I have been an SDET on VS2010. The real problem with VS in the past few years as it has move away from native. Everything is managed managed managed! We have left internal (Office) and external (lots and lots of folks) customers in the dust by only building decent managed tools. That's my opinion.

What I really wanted to address was performance. There was indeed a huge push after Beta2 to improve perf. At the end of the day 2010 is still slower than 2008 and eats up more memory. I guess that's an accomplishment if sucking less is considered good. But will people upgrade? I don't know. Hopefully because that's good for me. But I am pessimistic and currently looking for other employment as the experience was that bad.

Why was it so bad? Well I feel most of the pain points could have been avoided. Management ignored a lot of the performance/memory issues last summer and had an attitude of ship ship ship with NO regard for quality. I was responsible for reporting an overall grade for the areas I tested. I reported them to be poor and fair. I also listed all the known bugs in those areas. Management boosted the rating to good on the basis that memory and performance bugs don't count.

Let me reemphasize that last point: I was told memory and performance bugs DID NOT count towards the overall quality rating of the product. This was management's attitude through most of the product cycle. We could never give an honest assessment of the product because they would challenge it and argue that it wasn't that bad.

I believe their reason was simple for doing this. They need to take the data to their managers and up the chain of command to paint a pretty picture of how great the product is. Even if that's an illusion. Basically telling the truth of product quality is punished. If I lied about quality managers would be ecstatic about how a great a job we did. If I told the truth about quality instead of thanking me for figuring out where customer pain points might be the attitude was ALWAYS this should have been found earlier. The problem is many times I did find it earlier and they did their best to bury it before the next meeting they were heading off too report product quality.

I will say this. IntelliSense is much better for C++. Even though other native tools might suck. IntelliSense teams seems to be doing a pretty good job. It has always has been my favorite VS feature.

As an insider I think those numbers also include bonuses for Microsoft numbers at least. And being on the inside we are told we are paid on average more than industry. That may be true at face value but I really think we need to compare ourselves like this post does: to those we directly compete with.

But money is silly. The best thing MS can do is create an environment in which Senior Developers don't leave. But the only way MS knows how to do that is with $$$. Which is not sustainable.