[Thoughts] New controller for my X10 setup

While I was getting a special PCI riser for the new Netlantis server, I've decided to get a 8" LCD Touchscreen (from Linitx).
This product is really nice and easy to setup under Linux.

The idea for this screen is to replace my current X10 control, which only allows me to power on or off everything.

I created an image and a very simple script which calls heyu...

Here's the image.


And the screen, once installed and running.

[Thoughts] The stradivarius of Headphones

Notice the capital H to Headphones here, we're talking about high quality (and price) audio.



The Beyerdynamic DT 770 ed. 2005 is not just another headphone, to me it's an experience (wow, that sounds like crappy oldschool marketing). Let me clarify...
With this headphones, even the worst piece of 128k MP3 sounds like your getting a personal visit from the artist and his band. An experience you can only experience with very expensive loudspeakers, like these :

Sonus Faber Stradivari Homage :



Monitor Audio Platinum PL300 :

[SYSCTL] New Cyberabuse Whois 5.5 is out

This new version corrects a few minor bug and improves a few things.

As always, this new version is the best available, anyone using the cyberabuse whois should upgrade.

5.4 was released in 2006, so this is a long awaited release.

I'll detail the ChangeLog right here :

- fix Makefile for Win32
at least on the latest cygwin version, the current "make sqlite" didn't work, now it does
- gcc 2 compat issue
well, this one was introduced with the "make mta" thing
- get rid of that lame and big ascii art
- default TLDs libs have been updated
- TLD includes generator has been updated
- removed support for sqlite 2.x, upgrade to 3.x
- fixed a small bug in the ARIN cache system
honestly I don't remember what this bug was... it may have been fixed in 2006 or 2007
- fixed typo in INSTALL file for "make server"
- added better checks for netname and inetnum for KRNIC
- various improvement for LACNIC
- added "make mta" for those who use zcw to add abuse headers at the MTA level
that's not such a good idea... but I'll let people choose what they do with their mta
- added "make sqlite-spam" for those who use zcw to report spam
zcw was made to report abuse/security issues
this is probably the most awaited feature... I must say I'm not really a fan of this one and I don't know if it's going to be more efficient than the classic (sqlite) version...
- make strstr case insensitive when the searched string is lowercase
shouldn't change anything (email search), but who knows

I'm happy that this tool is still in use... it's now 5 years old !

[Thoughts] Hi-Fi upgrade

Big upgrade of my hi-fi system.

My Inter-M (Inkel) amplifier has been replaced by a Cambridge Audio 840A(v2).

Sound is really getting better, I don't miss the very dynamic sound of the Inter-M.

Long time listening of this system is now a much better pleasure and stereo image has been really boosted...

[SYSCTL] Lighttpd separate Error Log for 404's

You might have seen that my lighttpd conf doesn't include an access log... in fact I don't even load mod_accesslog.

Why would I want or even need to log access to my static content... you might need it, but I don't. Also, since I run multiple workers, even if I wanted to log server accesses, the log file would be broken (logs + multiple workers are not handled well at the moment).

However, if I don't care about 200s, I do need to get my 404 errors logged...
Apache logs 404s and > 400 in the error_log file, but Lighty doesn't (it uses access.log for that). However there's a nice feature called "server.error-handler-404" to bypass this issue.

This is how I use it... in the conf file :
server.error-handler-404 = "/errors/log_errors.x"
static-file.exclude-extensions = ( ".x" )
fastcgi.server = ( "/your_noatime_.../static/errors/log_errors.x" => ((
"bin-path" => "/usr/local/bin/log_errors.x",
"host" => "127.0.0.1",
"port" => 8200, # (why not...)
"min-procs" => 1,
"max-procs" => 1,
"check-local" => "disable"
))
)

touch /your_noatime_mounted_partition/static/errors/log_errors.x

Install FastCGI dev lib from :
http://www.fastcgi.com/dist/fcgi.tar.gz

Then edit log_errors.c :
#include <time.h>
#include <stdio.h>
#include "fcgi_config.h"
#include "fcgi_stdio.h"
#include <stdlib.h>
int main () {
FCGI_FILE *err = FCGI_fopen("/your/path/to/error.log","a");
if(!err) return 1;
while (FCGI_Accept() >= 0) {
time_t rawtime;
char buf [80];
time(&rawtime);
strftime(buf,80,"%Y-%m-%d %H:%M:%S",localtime(&(rawtime)));
char *uri = getenv("REQUEST_URI");
FCGI_fprintf(err,"[%s] 404 : %s\n",buf,uri);
// FCGI_fflush(err); - if low traffic and realtime logs desired
printf("Content-type: text/html\r\nRefresh: 0;url=/nice_message.html\r\n\r\n");
}
FCGI_fclose(err);
return 0;
}
// EOF

Then :
gcc -I/usr/local/include -L/usr/local/lib -Wall -O2 -o log_errors.x log_errors.c -lfcgi
mv log_errors.x /usr/local/bin/log_errors.x

And that's it...
Oh, this is not the most beautiful piece of code you can find... but it does the job.
I'll be happy to get better code if you write it...

[SYSCTL] Testing Lighttpd

In need for a server that would serve only static content (small files), but at the highest rate possible, I've tested Nginx, Lighty and Apache.

While Apache is clearly out of the game, Nginx wasn't that bad, however lighty 1.4.19 clearly achieved the best performance (even better than 1.5.0 with sendfile-aio) and was much better documented.

Since I thought I was really wasting my time testing different parameters (which many other probably already did) to achieve the best performance, I'm posting my config file here in hope it will be useful.

server.document-root = "/your_noatime_mounted_partition/static"
server.errorlog = "/your_noatime_.../logs/lighttpd.error.log"
server.event-handler = "linux-sysepoll"
server.network-backend = "linux-sendfile"
server.max-fds = 8192
server.max-connections = 4096
server.stat-cache-engine = "simple"
server.max-worker = 4 # (= number of CPU)
server.max-read-idle = 60
server.max-write-idle = 360
server.max-keep-alive-requests = 1024
server.max-keep-alive-idle = 16
server.bind = "127.0.0.1"
server.port = 80
dir-listing.activate = "disable"
etag.use-inode = "enable"
etag.use-mtime = "enable"
static-file.etags = "enable"
server.username = "your_favorite_anonymous_user"
server.groupname = "your_favorite_anonymous_group"
mimetype.assign = (
".swf" => "application/x-shockwave-flash",
".gif" => "image/gif",
".jpg" => "image/jpeg",
".jpeg" => "image/jpeg",
".png" => "image/png"
)

You can then use mod_mem_cache if you want even better performance...

[Netlantis] From Netlantis to FRNOG Tools

Some of my BGP related projects will now exist under 2 different "brands" :
- Netlantis
- FRnOG Tools

Here's the logo :

This might be temporary and we can only hope for the Netlantis projet's revival in a few months (years).

[SYSCTL] FreeBSD 7.0

One of my (FreeBSD) servers has been up for more than 3 years without the need for a reboot...
# w
8:38AM up 1138 days, 7:05, 1 user, load averages: 1.05, 1.35, 1.37
This one is serving something like 100 GB of data per day, the box is from 2000, a dual P3 Xeon 500 Mhz with 384 MB RAM...

This makes you wonder why the trend for switching from Windows to *nix hasn't been faster :)

Though it's nice to see so many devices running a *nix (mostly Linux) around us everyday. When Linux Phones will come to the mass market, people will be surprised to have a stable smartphone again (I have to soft-reboot my Win Mobile 2005 at least every 2-3 weeks). Though I really can't see why companies still stick with Linux... since the license (GPL) is clearly not the most flexible, compared to the BSD one, for example.


The very new FreeBSD 7.0 is available for download since a few days and runs on ARM CPUs (those used in mobile devices). NetBSD runs on ARM since a few years... Strange to see how some so tactical/important decisions are made.

Even though I'm still a FreeBSD lover, I mostly install CentOS (RedHat Entreprise clone) in production environment, because to me, it's simply the most production-ready *nix available to date.

[Netlantis] RANX beta

You can try RANX beta, HERE.

The RANX Graph feature is the one to be tested, the rest of the website is simple XHTML and should work on most modern browsers.

For this website, I've used several open-source libraries :
menus (which I've enhanced) : Dynamic Drive CSS Library,
RANX graph drawing library from Walter Zorn
(you may wonder why I've used a javascritp/DHTML lib to draw shapes dynamically... the answer is quite simple... I don't know Flash and I know several hardcore-GNUs who don't have it installed...),
the data for the RANX Graph are in AJAX/JSON format.

There's still a bit of work and optimization to do, but I must admit I like it already :) It's been really interesting working with these technologies that make 2.0 exist (namely, DHTML/CSS, XHTML, AJAX).

Don't hesitate to report bugs you can find.

[Thoughts] End-of-transmission - Changing Job

For those who used to contact me through my work e-mail pbourcier _at_ citali.com, this e-mail address won't be working anymore since yesterday was my last day at this company.

I'll be working for Criteo from now on... more stories on my new 2.0 work-life later :)