stgarrity

I left MSFT at the end of 2008 after ~3.5yrs (search, Mesh, Azure), took 6 months off, and started my own company. In the processing of recruiting & fundraising, the experience at MSFT was almost always a good thing (people respect it)--but I got a couple of comments from investors saying "it's a good thing you left when you did, > 5 years is a red flag for us".

I also have the same reaction hiring people now...a few years means I'll always interview them, MSFT's selection process is generally good. But after about 3-4 years, I've found most people (even the ones who express an incredible, proactive desire to "do a startup") never actually follow through on the decision. I think this is what Joel was referring to as the $300k/yr treadmill.

Finally...a couple people pointed out that the skill set you develop at MSFT is of marginal utility externally in specific settings...the specific setting that most HN readers assume is Silicon Valley. :) Assuming that's your goal as well, .Net-specific and Windows-specific experience isn't highly transferrable to LAMP down here. Obviously good developers can pick up new languages pretty quickly, but you are missing some depth there. And clearly the MSFT process isn't going to be helpful in most companies < 10k people, or honestly most web companies (and I worked on services there :) ).

Long answer short--I think 5 years is longer than you need or want if you want to go elsewhere, especially a startup...I'd vote for 3 years.

We're about to move ahead with Braintree and there's no multi-year contract anymore...the fees are a bit high but so far the integration has been super easy, and I like the lack of needing to deal with PCI-DSS since they do everything on their side (but still within your interface). I figure it's better to move ahead fast with the implementation and optimize fees later once we've proved the model...at least we can get our data out down the road if we need to.

why not just throw a hash function in the middle? technically you're not increasing the search space for a brute-force attack (but that's not the goal), but you are probably slightly increasing entropy. and most of all, you avoid bad PR like this. :)

that said, it's probably still quite expensive and this PR isn't that bad.

I do something similar, largely because enough of the stuff we do requires two-way interaction with other services that I need a public IP for callbacks, etc.

Sucks when you have flaky connectivity, though, and I've been gradually investing in stubbing out other services so I can "run" them locally/offline.

I also prefer a desktop so I can use more than 2 screens (my current config is 4, although the 4th is "just because"--3 is really helpful).

Clara & I met socially in college, partnered up for a compilers course (CS143), then did a fellowship together during our masters. Went our separate ways for 4 years after college (me to Seattle, she stayed in SF) but stayed close friends & traveled together, both for work & play.

We both quit our jobs this year to start something together...we've been talking about it for long enough it was about time. :)

Tangential to the article, but very related to the overall debate, I thought HN might particularly appreciate Burt Rutan's presentation from Oshkosh '09 on climate change.

He basically starts off "I'm not a climatologist, I've a flight test engineer--but I look at data for a living. Here's my thoughts on all the data I've seen about this debate."

http://rps3.com/Pages/Burt_Rutan_on_Climate_Change.htm