text stringlengths 0 1.62k |
|---|
**Break:** \[01:04:30.07\] |
**Jerod Santo:** So there were a lot of changes that went into 3.0 around the modularity, and yet I think 3.1 was probably a bigger deal as far as upgrade trouble. I think GitHub famously stayed on the 2.3 (I think) branch for years because of difficulties upgrading. Most of that I think was around the asset pipeline. ... |
**David Heinemeier Hansson:** Sure. First of all, I don't think that the upgrade trouble was from the asset pipeline. The asset pipeline was an optional piece from day one; it still is. The difficulty was definitely from 3.0, because in 3.0 we just changed a bunch of APIs, and we had to to make those extendable points ... |
Some of the optimizations - they had API changes behind them, too. So we were changing a lot of internals, and when you change a lot of internals (and we had a lot of changes like that), it just becomes harder to upgrade, and especially it becomes harder to upgrade if you have a Rails 2.3 application where you've made ... |
The upgrade from 2.3 to 3.0 was not actually terrible if you had no extensions to the Rails framework, but they were brutal if you had deep extensions to the Rails framework, because so much of the internal implementation changed, and that's of course exactly what a lot of big shops had... Because Rails 2.3 just didn't... |
GitHub might have all sorts of extensions to let's say Active Record that dug deep into the bowels of how the query engine worked or something like that, and then if the query engine changes, then those didn't work anymore. So I think that was the core of it. But I think the asset pipeline is still a good story, becaus... |
Rails has a very long history of making extensions to the framework and meeting resistance from certain camps, that like "Oh, this is not Rails' responsibility. You should not address this. Somebody can just deal with it in Gems or elsewhere", and the asset pipeline was certainly one of those things. |
\[01:08:08.13\] I'm still a little fuzzy on what the actual core of the opposition was about, because at least as I saw it, the asset pipeline made it easier for us to deal with JavaScript and CSS in a structured manner, instead of just dumping everything into public. But I think the problem with the asset pipeline for... |
But there's that, that's the practical concern, and then I also think that there was that philosophical concern, and we had that every day. Almost every week there's somebody in a pull request saying "Why does Rails need to do this?" Well, because it's better. Because Rails is better when it also addresses this issue. ... |
That reminds me of another big schism we had in the Rails community, which was the adoption of Bundler. I think actually the adoption of Bundler came with Rails 3.0. |
**Jerod Santo:** I thought you were gonna say CoffeeScript. |
**David Heinemeier Hansson:** That too, but that just follows the same pattern, so to me it's less interesting. I think the Bundler one was even more interesting, because the parallels to the asset pipeline are even clearer. To me -- I think it was probably Yehuda that showed me this, or maybe it was in a conversation ... |
It's funny, because this was not just novice users of Rails that had this opinion. I had epic debates with Jeremy Kemper over Bundler. He did not see the underlying value in Bundler for a very long time. There were some philosophical differences there too, which goes back to the debate of UNIX versus integrated systems... |
**Jerod Santo:** Imports, like a bunch at the top. |
**David Heinemeier Hansson:** \[01:12:06.24\] Imports at the top of the file, and they would just be like pages long, and I thought "That is just retarded. That's not how we're gonna play this game over here", and we didn't... Like, today in Rails, most Rails applications do not call require very much. Everything is au... |
I think to a large extent the same is true for the asset pipeline. There are still some holdouts who have their own build pipes, but I think that the opposition moved on to a different spot, which is more around what should Rails do with client-side MVC, and how should it deal with that kind of stuff. Now their sort of... |
**Jerod Santo:** Right. The controversy nowadays, as you said, is on how Rails handles fat JavaScript clients, and how that plays nice in that ecosystem which is becoming more and more popular as time passes... So how does it play? How now and how in the future will it handle JavaScript? |
**David Heinemeier Hansson:** So there's sort of two approaches there. One is some introspection about how large the Rails tent should be. At various times, I've been flip-flopping back and forth on this issue - should the Rails tent be so large that Rails is still a great fit for API-only servers, where Rails is not a... |
**Jerod Santo:** \[laughs\] Right. |
**David Heinemeier Hansson:** Both sides can think that's a good term; that's one of those rare moments I think it's great. Anyway, the appreciation I've come to is that we share far more than we're different, so what if we don't generate the view in the same way? To me, in many ways, it's a bigger point, but it's rela... |
I can divorce that from my own personal opinions about how suitable or not suitable client-side MVC is for a large swathe of applications. I personally want to build my applications in that fashion. |
**Jerod Santo:** Right. |
**David Heinemeier Hansson:** \[01:15:53.09\] I think that's a key foundation of why Rails is such a big success, is because we've found a platform where many people can collaborate even if they disagree on some of the particulars. We're not blowing up the community just because there are different opinions on the valu... |
I think going forward, we're gonna make it even easier for people who want to say "Rails, I don't give a hoot what you have to--" Actually, let me restrain that - "I don't give a f\*\*k what you have to say about the view", which is often the level of contentiousness that we have in these debates. |
**Adam Stacoviak:** Well, we couldn't get through the whole show without that, right? |
**David Heinemeier Hansson:** No. "I don't give a f\*\*k about what you have to say on your opinions on the view, it's just... I'm not gonna follow that, and we can still be friends. We can still be friends. And we're gonna move towards that..." So the practical matter of that is perhaps less than the cultural matter. ... |
**Adam Stacoviak:** "I'll pass you the salsa", I love it. |
**Jerod Santo:** So what does that look like? Does it look like you flip a switch in a config and just a whole bunch of stuff gets turned off? |
**David Heinemeier Hansson:** Yes, I think so. I think it's like $ rails new myapp--api, or something like that, where it just does not include all those bits and bobs that relate to build logic and the generating of HTML, and the asset pipeline. Those things are just not relevant if you're treating Rails purely as an ... |
**Jerod Santo:** That sounds like a good idea to me. I've built Rails API servers and enjoyed using Active Record and Action Controller, and just not having a view layer rendering JSON; it's nice, so I'm glad to see... I used the Rails API project, which is a slimmed down version, so it's nice to see a lot of stuff is ... |
**David Heinemeier Hansson:** That's basically the foundation, that's the spike. |
**Adam Stacoviak:** So maybe we can pause for just a second... If someone out there is working on a project that does what Merb did in the past, which maybe now has come to the table with the same opinion as you, they're something different that's an offshoot of Rails or a forked version of Rails that is better for API... |
**David Heinemeier Hansson:** I think the funny part is that the offshoot that we have is the Rails API project, which Yehuda and a bunch of other... |
**Jerod Santo:** Steve Klabnik... |
**David Heinemeier Hansson:** Yeah, Steve Klabnik and a bunch of others were involved in... That's probably gonna be a large part of the foundation of it. I wanna put a little bit of work into a serialization project that they've been using. One of the points of disagreement has been JBuilder versus... |
**Jerod Santo:** Active Model Serializers? |
**David Heinemeier Hansson:** Yeah, Active Model Serializer. I think there's values to both of them. I think JBuilder's probably a better fit when an API is just one of the things that you're doing, along with everything else, and then I think the serialization project makes a lot of sense when you're just building an ... |
Again, there's no underlying fundamental philosophical differences here, and I think that that's a key realization that you have to make to be able to have progress and stay in the same tent. |
**Jerod Santo:** \[01:20:01.06\] That's great. So is that Rails 5 stuff, or is that beyond Rails 5? |
**David Heinemeier Hansson:** I'm hoping it's Rails 5, yup. |
**Jerod Santo:** Awesome. What else is Rails 5? I saw on Twitter you were mentioning native websocket support. Is that...? |
**David Heinemeier Hansson:** Yup. So I'm working on a bunch of new stuff for new ideas in Basecamp, and a ton of stuff is spilling out from that, and I'm super duper excited about putting that into Rails 5. Rails 5 I think in terms of breadth of features and what it's going to do, it's gonna be one of the biggest upgr... |
I mean, I've already shared a bunch of sort of the fundamental ideas, which is the notion of the hybrid app, where you have a native shell that works on a lot of web view HTML content that's being served straight from a Rails app, and it's basically just third, fourth and fifth generation of that stuff. |
**Adam Stacoviak:** And now, a word from our sponsor. |
**Break:** \[01:21:45.24\] |
**Adam Stacoviak:** RailsConf is 21st-23rd April this year. That's a couple months away. |
**David Heinemeier Hansson:** Yeah, it's pretty close; it's not that far away. It's good, because we're also just at that point now where it's not like Rails 5 is gonna be released at that point, but it'll give me just enough time to collect my thoughts on the matters at hand, and be able to present something cohesive.... |
**Adam Stacoviak:** \[01:24:10.17\] Well, if we're looking at the timeline of the releases, 4.0 was out June 25th, 2013, so how far -- do you often do that? Do you look back and like "Well, we've gotta release a new major release every two and a half years, three years", or do you just do it whenever it's ready? |
**David Heinemeier Hansson:** It's funny because it's pretty much yesterday's weather. We came up with the internal clock that says "We should make a new point release every six months or so, and we should make a major release every two years or so." And how did we come up with that? We did not just sit down and... |
**Adam Stacoviak:** Guess it...? |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.