Building Serverless Apps with Webtask.io
auth0.com
auth0.com
I've had a webtask running since February when I made it to interview for Auth0 (ironic). It's still running, firing off events and contactable via HTTP. So when exactly does it get turned off since it's obviously past its lifetime?
I still prefer not to trust a blackbox such as Webtask to run my code. Luckily this webtask I'm referencing does nothing more than fire off push notifications. However, I'm waiting for the code that relies on it to fail when Auth0 does end its life.
Unless you need the scaling potential of the "serverless" platform, or are completely unwilling to setup/maintain a simple $5 a month VPS, I don't see much benefit to the platform lockin of a specialized service, when there is a boring but predictable way to achieve the same thing, with the same (or possibly less effort).
Once PHP is set up the process seems like it would be:
1. write your page to accept the query parameters or post data and return the results (the page is the function essentially).
2. Use SQLite for storage (either per user or per "function" maybe?). Other storage options are also fine, but SQLite is pretty simple and flexible.
3. Deploy the .php file using scp or sftp to your /var/www/whatever directory.
At this point you're done and can call your "function" via http.
Instead, there is just code and nothing but the code.
I don't spend much time "maintaining" my server for this kind of setup. Once it's setup I basically ignore it, accept for uploading new code.
I suppose I do occassionally run the system update command. I don't run any other kinds of PHP applications (WordPress, Drupal, etc) on an environment that I use this way, so I don't feel the constant need to check for security vulnerabilities.
At that point, you have to intervene, troubleshoot, and fix the problem.
But if you had gone the Functions-as-a-Service route, you'd avoid such occurrences altogether. For some people, that's an appealing trade-off.
I feel like I have the skills to handle that, or to create redundancy if necessary for reliability, so that doesn't bother me, but I can see the appeal of ignoring all of that.
I still like the idea that I'm not trusting a "black box" to run my code. I know how it's set up and what it's doing. And if I add replication or load balancing, I also understand what's actually happening there. But you're right, once you get to that level, you're out of the realm of "simple".
Maybe I can still hide behind my earlier comment "unless you need the scalability..." to say that covered your caveats, but I'll admit I wasn't explicit about the considerations that guide the optimal use cases.
The authentication/authorization is another big one. Auth0 is a lot easier to work with compared to Amazon IAM and Cognito.
Lambda does have it's strengths in the AWS ecosystem though and also supports multiple languages while Webtasks are mainly written in Node (and C# if you use a wrapper).
Disclaimer: I work at Auth0 as a Tech Writer and wrote the above article.