Good on you to build a tool to scratch your itch!

As another poster highlighted, hosting other people's content is tricky. Its certainly not a business I would rather be in. I bet the management of this SaaS service is going to quickly take over your day job.

If you are just hosting your own content you can do it pretty cheaply using the AWS Serverless Image Handler stack:

https://aws.amazon.com/solutions/implementations/serverless-...

Oh, nice.

Mine isn’t image-based, but I rolled my own CDN using S3 as a backend, Lambda/DynamoDB for some metadata stuff, and Cloudfront as the CDN (so yeah, I didn’t roll my own CDN, I just implemented all the backend upload/admin stuff).

The great part is that it’s stupid cheap for a hobbyist, and Cloudfront has a generous free tier. I can also always swap that part out if it becomes too expensive.

Mind sharing your secret chicken adobo recipe?

Yeah I'd grab the URL and feed it to curl on the command line to start inspecting the HTTP traffic and start from there.

You're hired.

Great points.

I did have the thought about attending meetups - which is a good approach. My first reaction was just that I dont want people to think I'm leaching off of them - I want to give back (assuming they want what I have to offer).

Its funny because I feel I am quick to help others w/o any thought of retribution - but I dont expect the same from others.

Thanks for your thoughts!

Another solution is wal-e[1] which handles continuous archiving. It was built by the Heroku guys and as such is battle-tested.

I use wal-e myself and its indispensable and easy to use.

https://github.com/wal-e/wal-e

After reading its README.rst, it seems to me like wal-e only cares about backing up (and restoring) WAL files, not about synchronuous replication like the original article.

But it is very easy to combine wal-e and the built in streaming replication.

I wish I could +1 this a million times. I develop a Ruby gem for the QB API (Online & Desktop)[1] and I've come to realize their whole API is a mess. I've also worked with the Xero API and its much simpler but still lacks some key features like adding a discount to a line item on an invoice.

[1] https://github.com/ruckus/quickeebooks

A brother in arms!

Check out imgix.com

how does this compare with 0mq?

Did you look at beanstalk? It has both of the features you're looking for: the ability to set a delay and when a job experiences an error it will be placed in the "buried" queue (which is like your Dead Letter Queue)

http://kr.github.com/beanstalkd/

It has client libraries in most languages and also definitely won't require the use of polling.

I did browse the beanstalkd, but I didn’t try it. Since we were moving our infrastructure to aws, it was feasible for us to use aws simple queue service. And I’m using long polling (which will wait 20 seconds to reconnect again) and bulk sending to reduce the no of calls to SQS.

This looks great but how does it compare to FullCalendar? [1]

From what I can see there is zero documentation, an area which FullCalendar excels at.

It will be hard to use this plugin without any documentation, certainly in the areas of events, which data sources are supported, etc.

[1]: http://arshaw.com/fullcalendar/