Moved again

I moved again, to a new promised land...

• homepage: http://infofarmer.cenkes.org/
• wikiblog: http://blog.cenkes.org/

My posts and pages from other places are being migrated to the new home.
Everything is in beta state, but everyone is welcome!

On to new blogging platform

With all my recent ranting about blogs and wikis, it turns out quite naturally most of the problems I fancied had already been solved long ago. I'm switching to a TiddlyWiki to keep my personal stuff - hosted at my hard-earned piece of FreeBSD cluster web-space - http://people.FreeBSD.org/~sat/ - complete with an RSS feed, but not looking very nice as yet. As for a wiki, we've chosen TWiki at work. It's not perfect, but it's got much potential, so I'm happy for the moment. I'm not sure if I'll be copying/moving over posts from here to my Tiddly or just leave them to rot here, but you probably won't be seeing much new stuff here. Cheers!

Wiki Scalability

The main strength of wikipedia is centralized collaboration, which gets redundancy considerably lower. But the strength doesn't scale at all. Nowadays we've got redundancy all over the wiki, and no way to deal with it. A structure would help, and it will probably be introduced at some point, but the migration will not be an overnight process. A real solution would implement something like reusable objects everywhere, through heavy info modularization. We also have yet to see an implementation of a distributed wiki system. Google shows up ideas dating back years ago, but I haven't heard about any real-life solution. I would like to see a DNS-like system, fully distributed, but not redundant at the same time, working across all wiki flavors. The kind of redundancy we would want to see for data security (backup) should be achievable through much the same mechanism DNS does.

And Still Computing is Immature

So many concepts are sci-fi only because many basic, infrastructural technologies are in their infancy. We say p2p - and think file sharing. But there's so much to share! I really marvel at some older protocol stacks, like TCP/IP and DNS. If only we could apply the genius in their design to our modern shiny new technologies.

Reinvention

Is it good when a great new idea you come up with turns out to be old news? Of course it doesn't feel that great, but then if you were not pursuing some personal profit, all that matters is - the implementation has already been done for you. Not quite the way you wanted, not perfect, no, but it's there. And if we're talking open-source software, then it's there for the taking. The last few months I've been spending/wasting much of my cycles thinking about computer-aided collaboration. So, I thought, wikis are great when you want to make changes quickly, but they have limited functionality. Why not let wiki users edit the wiki itself so they can quickly improve it - and built on it then. It appears that's about what is called a structured wiki and a wiki application. And there are some successfull open implementations, TWiki for one. I'm going to study the subject for a while.

The vice of imaginary automation

Whatever you do, you should always try to automate whatever you can, as it will save a lifetime for you, keep your tasks fresh and interesting, and reduce the errors you start making when you do the same thing for a thousandth time on end. That's approximately how on of the basic principles of system administration goes. Automation is good. But what if you see a way to automate most of your routine, but not sure if you'll be able to reach that automation in a limited timeframe? For me that automation is A.I., the topic I've been haunted by for the last several years. Everybody knows A.I. can do many things for us. We've seen it in the movies, read about it in the books and dreamed about it countless times. By acknowledging that pursuing A.I. is a lost effort for most of us, we set a precedent and a comfortable excuse for suboptimal operation. So whenever I type the very same 10 commands I've been typing for the last several months, I should be thinking that a simple shell script will save me a minute each day, but instead I think, what the hell. It's only a minute a day, why should I spend another minute on a script that could get useless any time and yet another minute to remember to use it instead of the literal commands. And if I decide, like a good exec, on a fixed ratio - XX% of time spent on automation, (100-XX)% on everything else - I would probably be spending XX thinking about A.I. :-)

Why I don't like blogs all that much

Blogs encourage you to leave your ideas as they are right after you post them. That's not the way you can write something good. In fact I leave most of my ideas before I even post them because I don't consider my blog to be a suitable place to work on them. I like wikis for the way they enable you to keep the stuff you're working on at hand and available to others. Still, wikis have some way to go for them to be able to replace blogs; some features, like proper aggregation and river-of-news presentation could be borrowed from blogs, others, like distributed workflow, should be designed and implemented almost from the ground up. What I'd like to see in a couple of years is a wiki/blog engine, where I could work on my ideas, revisiting them over and over, and people could browse through them like in an ordinary wiki, but also get proper syndication of important changes.

Poisonous language

I know it's not rational, but every time I see some fine idea implemented in Java, I frown. A couple of the latest ones I abstained from porting are LanguageTool and BitTyrant. I have several Java machines installed on many of my boxes, I use some Java programs from time to time, I acknowledge that some other languages are also somewhat fat and slow, but I just can't encourage our users to use Java. And looking at the ports collection and other software repositories, tightly coupled with FOSS systems, I guess many people feel the same way.

Spearbeheaded packaging

Days ago the Free Standards Group formed this Packaging workgroup. I'm sorry, but I just can't force myself to love that effort yet. The way it all sounds makes me somewhat skeptic. The whole thing is backed up by such industry monsters as SAP AG, and the members of the workgroup seem to aim at bringing the windows packaging experience to Linux. As porters/packagers we'll have to keep an eye on them because the system they are rallying for, named Autopackage, can really be the next ubiquitous package management solution, replacing both rpm and dpkg.

Footshots #1

Ever tried to "chown :group file" under BSD or Linux? Guess what, it just changes the group. Under Solaris it also resets the user. Imagine me trying to change the group recursively on a whole lot of www data, configuration files, databases, logs, lockfiles - all in one go and all on one of our production boxes. I love Sun.

A showdown of the elements

Today we found out that one of our server rooms had both heating and cooling working at full throttle. The problem is that the heater is a wall-mounted radiator attached to water-heated pipes. It's not that easy to remove it. We're filing a request to stop the idiocy, but chances are they won't get to it for weeks. Meanwhile we'll be watching two industrial air conditioners fight the heat with all the hardware in between.

Unicodifying FreeBSD

Actually :charset=UTF-8: and :lang=en_US.UTF-8: in your login.conf(5) (and don't forget about cap_mkdb(1)) are not quite enough. If you use a graphical login manager, like slim, it won't necessarily call login(1) and you won't get your environment variables. "export LANG=en_US.UTF-8 MM_CHARSET=UTF-8" in rc.conf(5) preconfigures environment for every rc(8) script and slim starts already in proper locale. Granted, it's hacky, but I'm not sure I know a better way. I've seen some issues with login.conf.db being regenerated automatically, but I still have to trace them.

Wave(let)?s

A friend told me about wavelets a couple of days ago, and how industry moves to embrace them in various compression technologies. The wikipedia article told me I would probably need 6 to 12 months of hardcore maths studies to understand the theory behind wavelets, but based on my impressions, I wouldn't bother right now. I still think natural intelligence is modeled around simple waves. More to be said at another time.

adjkerntz

What time zone is it anyway? Each day makes it harder and harder to tell. Today I listened to some classics with my eyes closed for three hours on end. That kind of tricked my brain into thinking I've slept enough. It somehow learned about the trick a bit later. MusicBrainz is doing something nasty to my brain, with it's huge and buggy Picard Tagger monster. Something impossible is going on.

Music Players Are Naughty

In pursuit for perfection I have always had a few players installed on my boxes. Apart from a bunch of console-based ones, I've used xmms, both generations of beep-media-player, audacious, and lately Amarok and Listen. My guess is I'll have to wait for about 3-5 years to see MusicBrainz integration that works, with an open audio hashing technology in use, a Last.fm that doesn't push AJAX to the limit, a good client-server model, somewhat alike what mldonkey is doing and a few other nice things. Meanwhile, I'll probably keep porting new players to FreeBSD and see what users have to say.

On the Go

So this was supposed to be a post made while I was on a bus, praising Opera Mini and how mobile browsing was approaching something it should have been for a long time now, but ironically Opera screwed up the http post request, sending one text field instead of both of them. So much for my grateful attitude.

BSD-BSD

Tonight I met a bunch of OpenBSD people. In spite of what you might think, I had a great deal of fun. Wim's attitude towards other BSD systems is not very generous, but when it comes down to eye-to-eye meetings, OpenBSD guys are very outgoing. And once connections are made, you can build on that in virtual worlds.

Sisters

Yesterday night Wim Vandeputte took me and Anton Karpov to a Sisters of Mercy gig. It was great; I couldn't make most of the lyrics out, though. Now that I downloaded some bootlegs and listened to the songs with texts before my eyes, my experience is complete. They rock! The funny thing is the Live in Germany bootleg from 1990 sounds almost exactly like the gig did. I guess Andrew Eldritch doesn't age with time.

A Comeback

It's been almost two years since I started this blog, not that I've been very active around here. Anyway, I think it's high time I started posting something, so get ready to consume.

The Beginning

So this is it, I'm finally in the blogging thing!