This is a post about Jobson, an application I developed and recently got permission to open-source along with its UI.

Do any of these problems sound familiar to you?:

I’d like my application to have a UI.

I want to trace my application’s usage.

I want to share my application.

They’re very common problems to have, and are almost always solved by building a UI in a framework (e.g. Qt, WPF), sprinkling on usage metrics, and packaging everything into a fat executable. In this post, I’d like to present an alternative.

That development approach is becoming challenging to do well nowadays because clients are more likely to use a mixture of OSs and 3rd-party resources. Therefore, new projects tend to need to choose between several options:

• Develop a CLI application. The easiest thing to develop in many situations. However, they can be challenging to use and have the usual client-side distribution problems (installers, etc.).

• Develop a UI application. Easier to use use. However, they’re usually less flexible for developers to use and are more challenging to develop.

• Develop a web API. Has similar flexibility to client-side CLI applications but doesn’t have the distribution issues. However, web APIs are challenging to develop because you need to deal with networking, sysadmin, setting up a domain, authentication, etc.

• Develop a full (API + UI) webapp. Easy to use and flexible (because of the API). However, webapps have all the complexities of the above and more: you effectively need to develop a full web API and a working UI.

When scoped in this (biased) way it’s clear that webapps are ideal vehicles for delivering the full “product” but CLI applications are ideal for ease-of development and flexibility.

It’s clear that developing a method for turning CLI applications into webapps would be valuable. It would enable developers to rapidly develop and roll out applications. That’s what I explored with Jobson: a web server that turns CLI applications into webapps.

Jobson’s Approach

Jobson has now been working in a production environment for a few months now and has proven to be an effective vehicle for delivering new platforms. However it cannot turn any application into a webapp. That would be tough: the potential “inputs” (applications) and “outputs” (webapps) are far too varied.

Jobson’s primary limitation is that the only applications it handles are batch applications that: a) start, b) take typical inputs, c) write typical outputs, and d) end. Most applications follow this model (e.g.echo, gcc, nmap, and ls).

With that limitation in place, Jobson could then be designed with simple principles in mind:

• Input applications are explained to Jobson via “specs” (at the moment, YAML files). The applications themselves should not need to be changed in order to suit Jobson.

• Jobson uses the specs to host a standard HTTP REST API which can be used by all general programming languages and utilities (e.g. curl jobson-server/v1/jobs)

• The API is regular and predictable: job specs should always create the API you’d expect.

• Jobson’s authentication uses standard protocols (e.g. HTTP Basic) that can be integrated into a larger systems easily.

• (Default) persistence uses plaintext files on the filesystem, which allows for transparent debugging and administration.

• Persistence can be swapped for other implementations (e.g. a MongoDB implementation)

• Job execution itself uses the underlying operating system’s fork(2), and exec(2) syscalls, which allows Jobson to run any application the host OS can run.

• By default, Jobson has no external dependencies beyond Java. It should be bootable from a barebones linux server with minimal configuration.

Overall, this design means that Jobson can webify almost any application very quickly. I could webify a major CLI tool in less time than it took to write this post, resulting in a web API and UI for that tool.

Why It’s Useful

Here are some typical development problems Jobson could help out with:

Problem: You’ve got a cool application idea you want to share

Without Jobson:

• Develop the idea
• Develop all the crap necessary to share that idea (download links, websites, databases, etc.)
• Share link

With Jobson:

• Develop the idea
• Write a Jobson spec
• Share link

Problem: You’ve got a toolchain (e.g. a pentesting toolchain) that has hard-to-remember commands. You want to streamline that toolchain such that you can run and queue multiple commands.

Without Jobson:

• Develop a UI around the tools, dealing with the usual development headaches
• Use the tool

With Jobson:

• Write a few small scripts around the tools (if necessary)
• Write a job spec
• Use the Jobson UI

Problem: You want to automate data requests at work, but there’s a lot of flexibility in the requests (FYI: this is why Jobson was made).

Without Jobson:

• Write an application to handle typical data requests. The application will have to deal with request queues, request structuring, etc.
• Find that end-users suddenly want to make different requests
• Redevelop, etc.

With Jobson:

• Write short scripts for handling a data request
• Host it with Jobson, which automatically deals with all the other stuff
• Find that end-users suddenly want to make different requests
• Add a new script + spec into Jobson

Overall, I believe this approach to developing a job system is extremely flexible and much easier to work with. Jobson abstracts all the typical job system faff (auth, APIs, queues, etc.) away from what’s actually important (an application that does something), resulting in a much cleaner system.

The next steps with Jobson are to streamline installation (by implementing installers), add a landing+documentation page, and start making tutorial videos for it. Watch this space =)