Hacker News new | ask | show | jobs
by jerf 5926 days ago
"Quite a disappointment, especially because I liked the idea."

The idea of building a music player on the Mozilla platform? That's what killed it, because it's actually a terrible idea.

Mozilla is responsible for screwing over a lot of developers by making glorious promises about the Mozilla "platform", but in fact to the best of my knowledge nobody has made it into anything "significant" that wasn't already in Netscape 4. And those who did went to great effort to do so (nvu in particular) and it is not at all clear they actually came out ahead vs. the other choices they had. (I know there's a handful of other dinky apps, but I'm not talking about whether someone bodged together a small local app in native XUL. The lofty rhetoric says that not only should Songbird be working better under Mozilla than it would in other places, it should be joined by the same kind of numerous "significant" apps that exist under QT or GTK.)

Historically, for the last six or seven years, trying to turn Mozilla into anything other than a browser has had exactly this result; bloated, buggy programs that barely work with a strong tailwind. And I'll tell you why: It's the platform. Mozilla is a browser. They made a lot of terrible platform choices that they managed to power through because it was a platform with basically one app (and Thunderbird off to the side), but it has not been able to get past that.

(RDF stores, where the fact that the RDF store is in RDF brings complexity, but no value for the complexity beyond other storage technologies. XPCOM, an abstraction layer with the same problem, nowhere near enough value brought for the complexity. XUL/XBL sort of eventually redeemed itself, but long after it should and I bet if you tried direct development in it for a "real app" it would still break your heart. And it still arguably fails the "not enough value for the complexity" test; lots of other things do what XBL does, much more simply and reliably. Note I know some of these have been supplanted; this is not a catalog of problems of today, but major errors the platform has made over time. And building something like a music app which is fundamentally multithreaded in Mozilla...? And the Mozilla of several years ago no less.)

Fortunately, it mostly doesn't matter, because it has done a pretty good job at being that one app. The only cost is all the developer hours sunk into developing on a "platform" where the only thing going for it is the grandiosity of the promises.

I say this partially because I was bitten, but not as bad as the Songbird developers.

1 comments

Komodo IDE and Komodo Edit by ActiveState is a pretty successful use of the Mozilla platform.
It used a "tweaked" Mozilla, I don't know if it still does. (I poked around but it doesn't say.)

http://trentmick.blogspot.com/2007/09/open-komodo-and-code_8...

which contains the interesting quote in the comments "Most current XUL Runner apps that I know about (those that get mentioned on planet.mozilla.org) tend to have their own patches against XUL Runner that haven't yet made it into the core sources to fix issues specific to their app."

Which was my experience with the platform too.

It is true that I have missed that, but you'd need to cite another twenty or thirty projects of a similar size before you'd catch up to the hype. (And I'd still question whether Komodo actually gets anything out of being on Mozilla vs. QT. Cost/benefits analysis of things on the Mozilla platform tend to miss the opportunity costs of having started out on a better-but-more-humble framework in the first place.)

(And while I won't hammer on it too hard, it would be nice if the app that is putatively on the platform actually ran on the platform instead of a modified platform. I don't have to go patching QT or GTK every time I want to make an app. If someone knows that it runs on XULRunner, great, please say so. All my googling is getting is people saying how nice it would be.)

"All my googling is getting is people saying how nice it would be."

Because that's all there is. It doesn't run on XULRunner. Getting it to do so continues to be a project goal, but it doesn't as of right now.

Maybe successful financially, but too slow for me to use.
I've never had any problems with slowness in Komodo IDE.