pnathan on software

One of the problems I've idly wanted to solve for a long time is the problem of automatic scheduling suggestion: you have a variety of things you want to do, but you don't know how to fit them all in.

Or, you have a variety of people wanting to find a meeting time, and have to poke around trying to find one that works. Blech!

I came up with a toy solver for that tonight using Prolog - it's over on a gist on my github account.

This particular solution aggressively leverages Prolog's unification facilities. First, I assert that a set of slots are available, then I have a query that determines available times.

In the available_times/3 predicate, it has three parameters - the first param (people being matched for) is a list pattern-matched into P1/PRest, then the timeslots. The head represents a person - we unify their availability, summoning the set of matching timeslots. Then, we move on down the list of people recursively with the Rest list. Eventually we terminate when the Rest list is nil. The key idea is that the Day & Time slots are being unified and the clauses in the query are *also* being unified. Therefore any solution for an individual person in the recursion chain is intersected with the solutions for the other people in the recursion chain; if there is an available timeslot found, it will be returned.

all_availability/2 is a mechanism to take a personlist and return a list of 3-tuples assigned to bag, using the findall/3 predicate.

It would be nifty to tweak this so that it was boosted to the point where, e.g., a list of classes w/ timeslots could be fed to it, then a set of possible class schedules spat out. Add on a prioritization scheme (must/may), a decent UI (Prolog's native CLI is awful), and this would be very useful.

My major upcoming (side) project is Critters, a game I'm writing. It's design for mobile phones - specifically, the Firefox OS phone. Being as FirefoxOS is HTML5+JS for a front-end, this means that it will work on arbitrary browsers (which are relatively modern). The basic idea is a virtual pet game, but I plan to make it quite a lot more. In particular, I plan to build certain AI features in over time. Why is that? Well, I noticed that, hey, a lot of mobile phone games are kind of uninteresting; not much intellectual enjoyment.

So Critters is an exploration into what that will take. How much fun can you pack into a virtual pet game, anyway?

Well, the initial fun is going to be driven by Ridiculously Photogenic Pets. Primarily, my cat Amos. You'll be able to play with a virtual Amos at first. For someone who grew up playing Warcraft 2, Unreal Tournement, etc, this is pretty lane. So...

After the initial version is viable and spinning along, learned behavior will be the next key upgrade I'll make: how you interact with your pet will matter over time. Poke your cat? Eventually he'll be upset at you and not purr (& do other things)! The overall learned behavior ideal is to have behavior I never dreamt up showing up in your pet's actions. Pretty cool IMO.

Prosaically, the business model is going to be up-front payment + in-app purchase for add-ons for your pet(s). Fundamentally, I want to ensure that my user's interests for giving me money are aligned with me getting them a better game experience. Ads are such a poor experience in my experience!

On the sheer geek front, my plan is to publish the API and have an official specification. The official spec will be encoded as a Haskell CLI program. This way iOS, Android, and other applications can be created if someone really wants to make them.

Looking at ObjectiveC today. I don't think it's a bad language. A bit weird. The combination of run-time dispatch + compilation is a strange one. I gather that it was an originally a C preprocessor. This shows in odd ways, especially with the reference counter system. The function declaration/call syntax is, IMO, a bit of a disaster. Square brackets, colons, etc. I keep reading the square brackets as Lispy brackets, where the function comes first, rather than Smalltalk brackets, where the object comes first.

Semantically I fear that dereferencing a nil pointer will be a big deal; run-time type errors are something I expect will become a big deal in my future.

Some fooling with bios hacking:

http://www.mitre.org/capabilities/cybersecurity/overview/cybersecurity-blog/copernicus-question-your-assumptions-about

...and that's not even interesting part. Seems to have a BIOS hypervisor, SDR functionality that bridges air gaps, wifi card removed.

— dragosr (@dragosr) @ragosr

note that dell & others make servers with out-of-band control (a.k.a. backdoor for malicious users) as a mechanism to automate command and control of datacenters. One person's efficiency-making tool is another's backdoor.

Things I need.

• Need distributed document store, encrypted.
• Need easy encrypting file editing
• Need easy key management/web-of-trust