Hacker News new | past | comments | ask | show | jobs | submit login

Funny, I was just thinking about this and Steve Blank.

Early at IMVU we really struggled with our crash rate. It was a 3D desktop application for _mainstream computers_. This was back before Windows 7 and everyone had terrible onboard graphics chips with equivalently terrible drivers.

During a board of advisors meeting, our CEO presented our crash metrics and justified it by saying "We might not be able to improve these metrics much - we're one of the only applications shipping 3D graphics on almost every computer."

Steve Blank immediately ripped into him: "That's an engineer's justification! Figure it out!"

I felt pretty guilty because I was the one who'd planted that idea in our CEO's head just days before.

Of course there _were_ ways to improve application reliability in the presence of driver crashes. You just have to move the rendering code into another process like Chrome does, detect when it crashes, and restart. :)




> "That's an engineer's justification! Figure it out!"

I find that I like Steve Blank less and less.

You might as well yell at your lawyers, your accountants or a physicist. Some things are.

The problem is that, once in a while, a professional opinion is wrong. So the Blanks of the world remember only the occasions where they bullied someone and lucked out -- then assume their Non-Excusing was the magic ingredient.


You can do the same on Android: the stagefright client libraries are terribly buggy, and sometimes crash if you get the timing wrong or tickle the demuxers the wrong way. It's possible (but undocumented) to send Surface instances across Binder to a separate service process and render there. There are no obsolete ideas.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: