Chris Oliver

If you're doing recurring donations, you'll need to store a User model of some sort so they can come back in and cancel it. You may not need passwords in that case, a secret token that you log them in with via email would work. You could make it work similar to a password reset token basically so they could manage their monthly donation via email.

You can generate a scaffold for it just like you did with the posts in the Rails class. You'll possibly want different fields, but you can create it the same way.

Those would be flags for bash, not cron itself. -l helps make sure your command for runs in a login shell so everything runs as expected and -c just says "hey we're passing the command through as an argument instead of running a script.

You shouldn't ever have to use sudo for this. If you do, you're going to end up with permissions errors later on. Homebrew might in order to change permissions for some of its own folders where you install apps from, but the rest of it shouldn't ever use sudo.

I can! In the meantime, check out Ryan Bates' episode on it: http://railscasts.com/episodes/406-public-activity A little old, but the gem hasn't changed much.

Yep that's it!

You can just associate all the records with the user_id. The only time you need Apartment is if the data should be kept private in their own databases for security reasons. You should be fine associating them to the user and making sure you don't load the other users's records.

Yeah, so you could have one Rails app that responds to all those domains and looks up the store like that. This would be easy to implement and manage (basically how Shopify works). I think you may possibly still need some sort of SSO in order to have the user logged in on the separate domains, but the code can all live in the single app.

Yep, that should basically be it, but the last step with the redirect, there will be something like a token that gets sent back that is verified.

The reason for that is because separate domains can only set cookies for their own. Subdomains can be included but not different domains which would be a security issue. Browsers won't let you set cookies between separate domains.

You'll have to get a token back that you can verify and set a cookie on the other domain as well to know that you're logged in on the other site. Basically you're making a microservice for authentication.

Not entirely sure what you'll need to do with the analytics, but that's definitely no fun.

Hey Thomas,

You'll definitely want some sort of single sign on process. Most of the results will probably come up on google if you search for that. I did a little searching and couldn't find anything really good. Basically you'll want a primary site that can handle the authentication and then send you back securely to the main sites so they can log you in. It's kind of similar to how OAuth works.

You may want to check out CAS whcih is a central authentication service. There's a devise plugin for it (https://github.com/nbudin/devise_cas_authenticatable) but I've never built anything with it. Pretty sure a project like 6 years ago that I worked on used a CAS but I wasn't around when they set it up. I should probably make some screencasts on this at some point. It seems really convoluted from all the tutorials but I'm sure it's simpler than they make it out to be.

I would say try your best to eager load as many of the subrecords as you can so that you're not querying as much. You may be able to cache that in the controller instead of the partial and save some query time that way.

You might check this out: http://brudtkuhl.com/securi...

I meant to mention this in the video so thanks for bringing it up! :)

Absolutely correct on that. Normally I would write a test for it, but this is not core functionality to Devise and tests for little tiny things like this don't really add much other than slowing down your test suite. There are still tests that make sure the generators still run correctly so it still works. If the functionality ever got reverted back to the original on accident, it wouldn't be a big deal. You could certainly add a test, but I'd argue how much value you actually get out of it in this case.

I think for resque, you'll need to pass in the tenant into the job as an extra parameter so that you can set the tenant in the beginning.

This is how apartment-sidekiq works https://github.com/influitive/apartment-sidekiq

Sidekiq is probably a better solution anymore than Resque. It's a lot faster and much better supported so if you haven't written too much Resque code, it might be worth switch to using the apartment-sidekiq gem with Sidekiq. It'll take care of passing in and switching to the tenant for you automatically.

For security, there's a few things you can do like setup a firewall and only open port 80 or 443 for web, setup fail2ban, and disable password authentication over SSH. You'll want to be careful not to lock yourself out of the server, but the recovery console can still let you in if you do. :) More stuff to checkout https://wiki.ubuntu.com/Bas...

Pingdom is probably the most used one, but there are a bunch if you search for monitoring. I use Pingdom's free service I think.

There's a cool little trick that if you change the youtube URL from "youtube.com" to "ssyoutube.com" it will redirect and give you a link to download the video like this:

http://en.savefrom.net/#url...

This might work. You'll have to test it inside the class methods to find out if the Apartment::Tenant is correctly set to verify that it's working. I imagine it would, but you can just print out the current tenant to verify it.

Possibly a better solution would be to loop through the companies inside the methods themselves instead. That would give you a bit more control over setting the tenants.

Hmm that sounds right. Are you able to share an example app that I could take a look at? Not sure I can give you any more pointers without fiddling with some code.

This a good question and is semi-complex so there are a ton of different approaches to it.

One of the ideas that's related to this (which isn't necessarily the right way to go) is Single Table Inheritance. It lets you define a singe "Service" table in your database and then have a "type" associated with it so that you could have ChefService, MusicianService, etc all stored in the same place. I would say this is a bad idea because you're going to have a lot of extra columns on the table that aren't associated with the service, but it's a step in thinking in the right direction.

There are a bunch of different ways you could do this. One option would be to create a bunch of different tables for these and just store them separately. Another would be to create a Service table, and have two other tables "RequiredFields" and "FieldAnswers". You could create the BaseService like normal, and then based upon the "type" column, you could reference the RequiredFields table and look up ones that match that type. You'd have a record for each like

RequiredField (service_type, name, format)

• service_type: "chef", name: "cuisine", format: "string"
• service_type: "chef", name: "dietary_type", format: "string"
• service_type: "chef", name: "max_people", format: "integer"
• service_type: "musician", name: "genre", format: "string"
• service_type: "musician", name: "instrument", format: "string"
• etc

And then this could generate a form dynamically based upon the fields and types. Then when they fill out the form, you can save their results into a FieldAnswers table that stores the RequiredField ID and the answer.

Another option would be to group all these into serialized hashes in a text field. It's somewhat limiting, but could also work.

Also this is where database like Mongo are a bit better suited because you can store a "Service" and have different types of attributes easily included. You don't have to define your data structure ahead of time, but it has plenty of other gotchas that you'll want to look into first. If you're down to learn all that, Mongo might be a good solution for you.