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

Working with multiple source control systems, multiple issue trackers, and multiple build systems has its challenges.

It's true that you don't know about all callers if you're working on open source software. There's no magic there; you need to think about backward compatibility. (On the other hand, if it's a library, your open source users can usually choose to delay upgrading until they're ready, so you can deprecate things.)

The main advantage for an open source project is that, though you don't know about all callers, you still have a pretty large (though biased) sample of them. If you want to know how people typically use your API's, it's pretty useful. Running all the internal tests (not just your own, but other people's apps and libraries) will find bugs that you wouldn't find otherwise.

There were changes I wouldn't have been confident making to GWT without those tests, and bugs that open source users never saw in stable releases because of them. On the other hand, there were also changes I didn't make at all because I couldn't figure out how to safely upgrade Google, or it didn't seem worth it.




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

Search: