The State of Meteor Part 2: What Happens Next

> Similar to how building houses is pretty straightforward.

Because we've been doing it for decades. The railroad wasn't straightforward to build when it first started. After a century there sure are lessons learnt. The conventional apartment building, or the conventional sky scraper, are not in the same league as some mega construction projects out there, so it's not always "straightforward".

Even though most startups and enterprises are solving very similar basic problems, the field itself is still quite young and open to innovation. In a century, perhaps building fairly complex social network with fairly high traffic might be as straightforward as setting up a Medium account, but we're not there. Like you said, there are many issues that are not easy problems to solve yet: dev ops, CI, etc.

Only semi-related...but why do you think Mithril is always so easily dismissed? I kind of get it: huge companies are working on and adopting React. But after using Mithril, I just can't figure out why it has failed to catch the attention of more people.

I remember that I found the documentation quite hard to read at the time I was checking it out.

I don't know if you are aware but there is now an official guide which goes way beyond the scope of the docs.

http://guide.meteor.com/

I'm no classical web developer, so I'm not be the best reference, but I used both a little bit for some realtime (mainly websocket driven) webapp for state visualization of remote embedded devices. I started with React, and I'm now on ng2 and happy with it (apart from some hopefully soon fixed bugs).

Imho both are ok. What I did not like that much with React was that the virtual DOM did not play nice all the time (strange effects can happen if you place non VDOM aware widgets inside it) and if you not take special care (keys) then reconcilation can produce a messed up representation. These things seem less problematic in ng2.

An unexpected very pleasant surprise for me was using the dependency injection mechanism in ng2 - I did not expect it do be useful because I never used something like this before. But now if any sub-sub-subcomponent in my view needs access to a status I don't need to pass it all the way down through unrelated components via props (like in React) but can directly inject it only in the using component. I also had good experiences with MVVM architecture from other projects with other frameworks, so I am happily using this approach also in ng2. For the model component I'm using service objects which are encapsulating the current state, exposing it through observables and are injected into the component (viewmodel) through DI. What I like here is that I can bind the view directly to the observables from the model with the help of async pipes and thereby get an auto-updating behavior with zero boilerplate code.

Apart from that there are obviously the major differences - angular has an inbuilt router, React not. But still I wouldn't think of Angular2 as something highly opinionated, you still have lots of flexibility.

As I am a big fan of typescript a factor for me was also that ng2 is typescript-first and I don't have to rely on possibly incomplete or 3rd party type definitions. But that might be more of a political factor, the basic typescript support for react is also good (it can even check the types of props in JSX - which currently not possible for ng2 templates) but many 3rd party libraries don't seem to provide type definitions.