Need a hand with tech consulting? I can help!
Learn more about how we can work via black.af .

The Idea Behind Wintermute

Giving some hindsight as to why I built Wintermute.

:pencil: by Jacky Alciné Jacky Alciné :book: an wintermute post :bookmark: wintermute , thoughts :clock7: written :eyeglasses: about 3 minutes, 698 words :link: Comments - 0 Mention(s) - Permalink

WINTERMUTE is the software project that propelled my knowledge of simpler programming paradigms like inheritance, polymorphisms and the likes over the last few years. I’ve taken the liberty of at times, blowing away the entire commit history just to start anew. Wintermute had a proper and here, I plan to expand on that.

Wintermute, when described to someone outside of the grinding gears of the tech multiverse, is an ambition to implement a computer network and interface similar to that of Iron Man’s J.A.R.V.I.S. The omnipresence, vast amount of semantic knowledge it holds and its ability to use both of those to provide context-relevant, if not crucial, information about the current scenario is the core essence of what’s to be captured by the Wintermute project.

Knowing Itself

Wintermute requires a means of walking over seemingly massive sets of data in mere moments. We already do this with current Web applications but how massive is Wintermute’s massive? Realistically, I can’t put it in a proportion for the purpose of having a working analogy but I’m going to go with the hunch that it’s extremely massive. What I can say is that the kind of data Wintermute would probably get on a bad day would be similar to a few engineers at Facebook running a service that’d check the number of times a person used a word of numerical significance in their account’s entire existence and that of their friends. A lexical- and graph-based search!

The kind of information that Wintermute would hopefully collect wouldn’t be saved in a classical relational database. Instead, each event and moment would be saved into a ontological database. Such a database would preserve context on each row or event and allow Wintermute to act in the scope of that event.

Communicating with Itself

As most networked applications are, Wintermute wouldn’t be just one instance of itself running on Joe Schmoe’s MacBook. What’s envisioned is that some sub-instances of Wintermute would run on a DIY kind of supercomputer, like the Parallela suite for USD$100. This could handle the work of the heavy number-crunching, the discovery and cataloging of new data and reindexing of the old. A robust, tried & true socket communication library would be more than needed for such a job.

Serializing messages into a format that we can quickly dispatch between processes has been more or less solved by using partial selections of representational data. Additionally, to do a cheap form of data integrity checking, including a checksum with each message allows for one to then challenge the actual integrity of provided data.

A formalization of what processes, or rather modules, are available on a machine wouldn’t be necessary for the sake of remote execution but they would be necessary for discovery and maintenance of a suite of processes in order to ensure that each running processes has what they need to operate. These modules could have exposed methods that a remote process could then invoke and await a reply, if said remote method has a return value.

With parallel processes, message serialization and remote procedure invocation, Wintermute can run across a vast amount of machines and provide the power it needs when it needs to!

Building Wintermute

The technical logic behind Wintermute is a bit weird. Initially, I tried working on the project as coming from a core executable meant to run on dedicated machines and build platform-dependent executables that then fed to the core executable to handle ‘raw Wintermute-esque’ work. This made less and less sense to me so I decided to wipe the slate clean and try a different approach.

The Library and the Driver

The initial version of Wintermute isn’t going to be an executable. Instead, it’ll be a library that handles things that Wintermute plans to abstract away. These would include plugin loading, remote procedure loading, process identification on the procedure pool and a bit more. As usually with most of my C++ projects, I’ll be using Qt for this, but it’ll be Qt4 for now until I can wrestle using Qt5 properly with CMake. This also means I’ll be focusing on making CMake support for Vim even better. It’ll be fun!