Happy to answer any questions about this, it is particularly cool because you really can run a SaaS service with "no servers" and still get to use your pick of providers (Sendgrid, Mailchimp, Twilio, etc.). It seems like a lot of fun, especially paired with Angular.js.
I think it is a question of efficiency, right? If some external middleware can provide more value in some columns than it offsets in other columns, it is usually a good bet. We're all just raising the bar for abstraction: first it was higher level languages, then it was libraries, then it was frameworks, and now we're talking about middleware (I might be missing a handful of steps in there).
Sometimes middleware isn't the answer, but sometimes it speeds things up so much that it is rather "cut off your nose to spite your face" to ignore it. The argument certainly doesn't apply only to middleware, of course. The same for the counter-arguments. So...
> Is it a good idea to build things without middleware?
As a person who have replicated Zapier kind of integration into our development tool, I don't find this news anything bug a paid advertisement. Anything with REST API can be integrated with Zapier, so Firebase can be integrated too. Looks silly to me.
Well, I am not saying people wont pay. I am saying making big post in TC every time some one adds an another API to Zapier is silly. Zapier is good idea, and I like the way they implemented.
We thought this one was particularly novel, so I disagree (though as a Zapier co-founder I am a bit biased!). We rarely submit API additions to HN unless it seems particularly suited for hackers, others in the community might though...
