Except that when an open source project dies, its code remains in full -- to study, take ideas from, or even potentially be revived at a later time.
When a language dies though, it's pretty much gone. Even if it's received a lot of attention from field linguists, it's near impossible to fully codify the grammar and document the nuances of any language.
Almost all the languages that we "lost" did not have a rich literature or sometimes even an alphabet.
The # of languages currently spoken is typically pegged somewhere between 3000-6000 and it's thought that we lose around one a week. Most have almost no documentation whatsoever, and even so a couple papers written in the 1970s aren't going to capture any potential unique aspects of a language let alone allow for reconstruction at a later point.
The analogy to code is way off.
The analogy to open source projects is spot on; while you might have some "emotional" attachment to a library/technology that you learned early on, you probably wouldn't make significant decisions based on that. I learned to program in VB5, yet I dropped it as soon as I learned Java because it was just an awful language (though I guess human languages don't really have notions of superiority or inferiority like some programming languages might).