Solipsism

Wed, 21 Jan 2009 00:35:11 GMT

Here I am again, back in Germany. I'll try and summarize my experiences with piman in the states, and I hope this post won't turn too much into a diary entry of a 14 year old.

As I wrote before in the last post, I was leaving three months ago to visit him, and I didn't quite know what to expect. I was looking forward to it on the one side, but also was very nervous and anxious about it. We knew each other for some time now, met online on IRC. But was it really such a good idea to see someone else, with potentially romantic interested on both sides, so relatively short after my break-up? My last relationship sure left its marks, but I was trying to see the journey as a visit to a friend, without strings attached. It should all just be hanging out with a ~~total weeaboo~~ ~~helpless otaku~~ cool guy.

After a few days in CA, I got to meet some of his friends, at a weekly get-together with watching anime. The shy person I am, I felt uncomfortable being around so many new faces, but in the end did okay. I went there again for one or two times, until I realized that watching a series starting from the middle isn't very entertaining for me. We frequently went out for dinner and I got to know some nice (bad)food places, like Mojo Burgers or Denny's. At first he was not very happy about my bad culinary tastes, but in the end I also liked it when we went to a sushi place. I tried eating most of it, but the only thing I really liked where the shrimp.

---
This draft is now a week old, and now I decided to just post it, even though its not even half finished :/

kfI - kein finanzielles Interesse

Fri, 17 Oct 2008 14:24:00 GMT

Sooo, here I am, IN AMERICA.

Here is what happened:

met piman on IRC
found him pretty cool for a while
light insecurity
gotten closer to each other
caused him life-influencing thoughts while on near-death-experience
heavy insecurity
got imported to america
heavy insecurity
met piman at San Jose Airport (he brought cookies, fucking awesome)
fucking ridiculous insecurity
staying at his apartment
having a great time
yet undisclosed positive emotions

I think piman is a pretty cool guy. eh buys me drseses and doesnt afraid of anything.

In A.D. 2008

Fri, 26 Sep 2008 16:23:49 GMT

I think this is a signal to write about what happen.

I finally set up vibrator-job the bomb for a more secure one (still not that secure though) at my aunts company. More pay, still under par though, but easy work so I know what I do. Also I can work from all my base again and thus have a rather free schedule. This months rent is paid, but next one is not belong to me. Since I might have no chance to survive I better make my time.

Zig, where are you when I need to take you off the most.

Sat, 06 Sep 2008 17:05:05 GMT

So, there I have another job. This time more regular and with more pay (not fulltime yet), still webdev though. The thing is, the other job is not gone, and I still need to finish something I started there. Now I have to manage two part-time jobs, which essentially are more than one full-time job, due to the working times that aren't set in stone for either job.
Deadline for one job's project is next Friday, the other one doesn't have one per sey.

And then I need to go to the job-centre applying for social money again on monday. I hate stress.

webcloggery

Wed, 13 Aug 2008 06:35:13 GMT

I haz a job now.

Mon, 21 Jul 2008 13:33:29 GMT

I'm doing pretty well lately, in regards to certain emotions. The rest of my life is still fucked in its own lack of motion, though.
I try to not think about anyting, since using my brain would be realizing what crap I'm in.

On the other hand, it always was easy to get out of any shit when I tried to change something. But why the hell should I do that.

Evil persons

Sun, 13 Jul 2008 05:48:41 GMT

I consider these persons evil:

Joe Wreschnig
Nou Sphabmixay

More to be added later, I suppose D:<

CURE

Mon, 08 Oct 2007 18:53:57 GMT

Contextual Universal Relation Engine

After managing to define how contexts should be encoded in a Pile, I am going to design the next step: CURE.

It is a layer upon the file system providing the relational space that is Pile. It represents the logical domains in Pile as the associated Relations as file tree structure, like a regular file system, and can be explored via ordinary file managers. I am going to implement it with FUSE. Here is the general idea:

Accessing the files or folders would be translated through CURE into the corresponding Relations and Domains. So the user sees only data and meta-data, not the actual Relations within Pile with which they are encoded.
Double-clicking an appointment could open the program responsible for creating it, like an appointment manager.
Also Folders (Domains) display the Domains they are connected to as sub folders within themselves. This makes a traversal across contexts possible for the user, without an extra search program. However, a specific search program would be far more flexible. It is important to note, that those sub folders are not hierarchical, but only relational. That means, one Folder (Domain) can (and is likely to) be in more than one Folder (Domain) at the same time.

Navigating through CURE is done via paths, like ordinary file systems, which I did not define yet.
A possible way is to structure the path as query to the Pile, like a SQL-query. For example:

cd cure/appointments/peter

would list all appointments where peter is involved.
Boolean operators could be supported to:

cd cure/appointments/peter/AND/mary/NOT/time/15:00

which would yield all appointments with peter and mary that aren't at 15:00.

Many possible ways to do that, but I'm not yet fixed on one particular way.
In other news, I've ,looked at other implementations of relational/non-hierarchical file systems. Here are nice ones:

RelFS - a linux user space file system using fuse and a relational database to store information about files. http://relfs.sourceforge.net/
LFS - this paradigm associates logical properties to files and parts of file, and logical deduction serves as a basis for navigation and querying. "Paths are formulas" http://lfs.irisa.fr/~pad/soft/LFSWEB/lfs.html
Tagsistant - a semantic filesystem for Linux kernels http://home.gna.org/tagfs/index.shtml
NHFS - a nonhierarchical file system http://rffr.de/nhfs
WinFS - a data storage and management system based on relational databases, developed by Microsoft http://en.wikipedia.org/wiki/Winfs

They all follow the same idea, pertaining meta-information with files, but they also are all the victim to container-dichotomy: they still store everything as pure data (with tags attached) and can't relate the between relations themselves. The idea for using paths as queries comes from the FUSE-based implementations of them, I think I could use that.

System Layers and Stacked Agents

Sun, 07 Oct 2007 02:12:15 GMT

I took a step back to have a greater perspective of the whole system we're going to design. So I thought a diagram depicting the general structure of layers would inspire me, and thats what I have done.

The two common layers today are the Application layer and the File System layer, in between is nothing, most of the time. What I did is packing another (two) level(s) of abstraction between them, to maintain the goals Pile offers: context.
I want the Pile space to behave like a virtual file system, where contexts are visible at once and from everywhere withing a Pile.

More coming soon...

Domains

Tue, 02 Oct 2007 21:12:24 GMT

Since yesterday I tried to work out contexts in Pile. I had one general idea on how to bring them into Pile, and that is with "Domains".

a Domain (-relation) is the normative root of every Relation in one specific context
the representations of Domain definitions in Pile are not worked out yet
a normative child of a Domain (-relation) is called a Domain Root
the associative parent of a Domain Root is a Relation in another Domain (or a Pile Root relation)
the most basic Domain is the Bit-Domain, which contains bits and bit composites (0, 1, 01, 11, 010, etc)
one "thing" can be in multiple Domains (one Relation representing that "thing" in each Domain it is in)
Relation Qualities are defined by the respective Domain

I'm low on words, so I will just slap up this diagram:

You can see the Domains rather clearly separated. Note that in this diagram, which is work in progress, the Domains are not connected to each other. But they are in my head, I just didn't draw it yet. In the very top you see the two Terminal Values, representing binary one and zero. The SD, System Definer, is there to form something like a starting point for the whole Pile. The two Tops for 1 and 0 connect to it in a normative manner. Then we have our first Domain, containing bits and loads of bits, up to 8-bit-pairs. The next Domain is the Byte-Domain, saying that eight consecutive bits can be interpreted as Bytes. That goes on with the Integer-Domain, which handles integer numbers consisting of a various number of Bytes. You get the idea, but is important to note, that those are not actually layers on top of each other, but rather spheres next to each other, not specifically ordered. They just happen to be logically related in a hierarchical manner.

Further down you see the three important Domains in this diagram: String, Time and Appointment. You see that the Hour and Minute-Relations in the Time-Domain are associative children of two Integer-Relations. This implicitly says, that a hour "can be stored/interpreted as" a number. Say we want to get the Data represented by that Hour-Relation, an agent would cross through the Integer-Domain in its traversal through the Pile, and know "ah, this Relation has traits from numbers". My idea there, is that the Agent applies transformation/conversion rules from the Integer-Domain to the Data it gathers, returning a meaningful value to the calling Domain. In the next diagram I draw, I will think about Domain-definitions and how that could actually work.

The bottommost Relation represents an appointment, happening at 14:00 with PETER.

I hope this is clear to at least someone. Anyway, I need to get back to my white board (the diagram which contains all diagrams I ever drew about Pile) and ponder Domains some more.

Current Goals

Mon, 01 Oct 2007 11:13:07 GMT

Pile has been working my mind since a long time (relative to my scope of "long", a 3/4 year). For that time I have been trying to get around the concept, and lately I came to a dead-end when it comes to understanding the possible solutions that Pile provides for different computation problems.

So in order to fully grasp the possibilities and to further develop my own Pile implementation, I needed use-cases which directed my thoughts in the right direction, specific to every use-case I would come up with. But the overall universality of the designs I might come up always were my major concern. I wanted it not only to make sense in one context, but also apply to all other contexts as well.

That being said, I am trying to come up with small real-world-application-ideas in Pile space, unifying them under the hood of a theoretical PileOS. So to validate a specific solution of an application implementation within Pile space, I also need to think about the impact it has on the general PileOS (and all other applications within it). This is clearly not an easy task, since contexts domains, by definition, intersect each other (not only in Pile, but also in the real world). For example, take the English language. The word "too", has different meanings, depending on context of the statement. To reflect those contexts in Pile, one needs to come up with a way to define contexts. And Pile agents would use and interpret the context according to that definition. I am breeding over a way to define logical contexts for some days now, and here are my thoughts so far.

agents need a definition of context to work with
contexts could be defined by the use of EBNF notation
how to define the level of abstraction an agent works on?
what meaning do intersections of context have in Pile, and how to work with them?
are there meta-agents that coordinate other agents?

So that are the things that worry me at the moment. And by the way, if someone doesn't know some terms, it is likely that I explained them in the first post.

About this journal

Sun, 30 Sep 2007 13:05:11 GMT

I welcome you and myself to my very own journal :)

It is about Pile and my time exploring it. Whether or not this journal will contain more than one or two entries... I'm not sure about that myself. We will just see.
The casual, non-tech-related reader won't find any interesting thoughts here, but I try to keep the thematics as simple and entertaining as possible. Be warned though, that I'm not a hobby-writer, entertainer nor a scientist. I am just a 20-year old girl who is interested in all stuff tech- and philosophy-related.

What is Pile?

Pile is a different, and relatively young, approach to data. By that I mean its storage, handling, interpretation and context. So, pretty much everything related to data. Oh, look, there is the first keyword already: "related". This is the foundation of the whole Pile-idea; instead of storing sequential data and/in its containers, Pile stores only its relations between them. Here is an example:

Conventional storage would store the string "PETER" as is, that means the storage looks like this:
[ P | E | T | E | R ]

It splits the string up into its characters and stores them one by one. Pretty convenient, isn't it?
Well, yes, but it has drawbacks:

Redundancy - because every character is stored as is, E is stored twice; therefore a text with 2000 occurrences of "PETER" would need to store it 2000 times
Loss of context - the characters themselves don't "know", that they are part of a whole, the word PETER; the only context they have is that they are stored in sequence; therefore there is no way to know from just the T in PETER, that it is part of PETER, even though it is contextually obvious!
Linear data volume - because there is redundancy, the amount of space it takes to store data, increases linear to the amount of data stored; straightforward but not optimal, is it?

The same string in Pile, would look like this:

More on this later...

Meanwhile, enjoy these diagrams:

(I am aware of Ralf Westphal's blog about pile - I needed to put down my thoughts, too, so I started this one)

First Test

Sun, 30 Sep 2007 10:55:41 GMT

First test entry to see how it looks.

Nothing special here.

(Really.)