Spectrum is now read-only. Learn more about the decision in our official announcement.


Our mission is to make cloud computing as easy and accessible as mobile computing. You can find our Next.js community here.


Thoughts on why Now 2.0 is kinda disappointing

November 11, 2018 at 9:14pm
The ZEIT community has a new home. This thread is preserved for historical purposes. The content of this conversation may be innaccurrate or out of date. Go to new community home →

Thoughts on why Now 2.0 is kinda disappointing

November 11, 2018 at 9:14pm
Hi, everyone!
First of all, I’m not hating on anything or anyone. Now 2.0 is an incredible product, but I have some thoughts I think many people will agree with
I’ve been a Now user for over a year. I came across it as a “Realtime Node.js deployments” platform, and I absolutely loved it when I first tried it
Then when the Docker support came and I couldn’t be more excited
The thing that attracted me about Now was the ability to quickly throw together a simple Express app or whatnot, type `now` and have it live in a few seconds. That simplicity is what got me (and I think many others) to fall in love with Now.
Then a few months ago comes the Serverless Docker beta announcement labeled “Cloud v2” with tons of awesome features. And then few months later Now 2.0 releases and it’s Lambdas.
I love the idea of serverless and lambdas, don’t get me wrong. But this model loses everything that made Now great in the first place. I can no longer just throw together a quick express app and deploy it, I need to re-think the way I build my entire application to fit serverless paradigm, while doing what I need in express would take me 2 minutes.
Which makes total sense as a “Global Serverless Platform”, but makes no sense as “Realtime Global Deployments” platform.
Essentially, Now 2.0 stops being a *deployment platform* and becomes an *application platform* along the likes of Firebase. Which is great, but not quite what we all signed up for
mentioned in another thread that apparently Zeit learned that serverless docker is bad during the beta, but there was nothing announced in that regard. There was no “beta summary” blog post, no thoughts on it vs. lambdas nothing at all.
All we got was “Hey, docker sucks, here’s the new thing”
Also I can’t help but feel that you guys rushed it out. Most of the examples don’t even work, there are errors all around, docs are unclear. It seems
majority of effort went into marketing pages…
Zeit people mentioned the tremendous scaling capabilities of lambdas and how it makes teams work better, all of which I understand, but I’m not a team, and I don’t build massive applications, so those benefits don’t really mean anything. I guess solo developers are not the target audience of Now 2.0
That said, I think as a serverless application platform I think Now 2.0 will be really great once all the kinks like cryptic errors and unclear docs are worked out.
But I (and I think a lot of Now users) don’t need an application platform. I need a deployment platform. I just want to deploy my docker 😔
Show previous messages

November 11, 2018 at 11:06pm
What about using puppeteer in our apps?
This is simply impossible to achieve unless we have a system that is aware of some "common runtimes". This is the "secret" behind the scalability, low cost and performance of Now 2.0.
This is why I don't write off docker in the future, but you will have to pay a cost. That cost could still not be in cold boot, if you pay a lot more money, but the cost has to be somewhere.
Is this "cost" any more expensive than the cost we had in the serverless docker beta?
The consequence is that those 220kb will be so fast to acquire and boot up in the future, that your system will scale infinitely, with incredible performance, and remain very cheap.
That said, I think as a serverless application platform I think Now 2.0 will be really great once all the kinks like cryptic errors and unclear docs are worked out.
We invested a tremendous amount of effort in improving our documentation, so this comment is a bit surprising.
This is a commitment you have from me, it's my current obsession. The best docs in the industry or bust.
I wouldn't mind having slower cold start performance, if it means I can put more than 5 MB into my image
that's the problem, however. "If everything is a priority, then nothing is a priority"
The vast majority of our users will want their customers and API calls to be resolved instantly.
However, we are aware of some use-cases where bootup times are not as important, like you correctly point out.
And we are looking forward to introducing APIs and configuration options to accommodate those.
By the way, I'm really glad you posted that screenshot!
Are there any additional "problems" with increasing the size limit besides bootup time?
That screenshot reflects exactly the tremendous advancement that we have made with Now 2.0.
We have stayed 100% true to our commitment of expressing your workloads in terms of fully standard protocols and APIs.
As I showed, I deployed express, all our Node.js functions use the Node.js standard API.
We have even accomplished this for Go and Python and all our other runtimes.
But... now.json.
And language flexibility is still possible. Some people in our community have already deployed Rust, Haskell, Elixir…
But why ditch docker?
We have all the tools in the works, in open source, to be able to export your app as a Docker container and local workflow.
If you use something like node:alpine, the image is still small enough for fast bootup
as I mentioned, there are tremendous scalability and cost tradeoffs. A Node.js container is 40mb vs 200kb.
Show more messages