Olivier's Blog - Internet

GeoIP 1.0.5 released

nospam@example.com (Olivier Hill) — Fri, 19 Dec 2008 15:02:09 -0500

I just released GeoIP 1.0.5 on PECL. An excerpt from the ChangeLog:

Small bug in phpinfo() when printing version number could crash PHP. (in 1.0.4)
Fix memleaks, bug #14851
Small patch for MacPorts by jhohle
Add geoip_time_zone_by_country_and_region() and geoip_region_name_by_code() (thanks to Sid Dunayer)
Add continent_code field for City Database, only if using GeoIP lib 1.4.3 or newer

Since most of the features are very dependent on the GeoIP Library version you're using, I do suggest you upgrade the installed library on your system before upgrading the PECL counter-part.

I have also upgraded the PHP documentation, which should appear on the mirrors by next week. I am now working on documenting the added functions, mostly that FIPS 10-4 thing in geoip_region_name_by_code().

One thing to note about this release though is that the time zones returned by geoip_time_zone_by_country_and_region() with library 1.4.5 don't seem to make any sense here. I have opened up a bug report on SourceForge for libgeoip, will see if it gets resolved in the next release.

In the meantime, happy IP Geo Localization with PHP!

All new blog!

nospam@example.com (Olivier Hill) — Fri, 23 May 2008 09:18:12 -0400

Please welcome... wait for it... my new blog! (you need to use Barney Stinson's voice for that phrase)

Ok, so what's the catch? It looks like the other one, what's different? In fact, what changed is the machine behind.

After many problems with ServerPronto, I decived to move everything back in Canada. My new hosting company is now iWeb, right here in Montreal. I wasn't sure at first, since iWeb is a little bit more expensive than others. After trying it, I am more than pleased with the service.

For example, when installing my dedicated server, I wanted to switch my Debian to Gentoo. I knew it was risky, but I needed my server to run Gentoo, mainly for seldom ebuild development. Everything ran smoothly, but I unfortunately made a typo in the network config file. Upon reboot, I couldn't access the machine and was stuck with a bricked server.

With a simple support ticket, I was able to get an IP KVM connected to my server quickly and interacted with it to bring the network back online. It might seem like a trivial event, but I am not sure what would have happened if it was hosted elsewhere. I would have probably ended up paying someone to type the commands on the shell, which is not always an easy task when the other person is more than a thousand kilometers away.

Since when PHPQuebec money driven?

nospam@example.com (Olivier Hill) — Wed, 05 Mar 2008 08:19:38 -0500

Although this article might seem like bashing, it asks a very legitimate question about how much the PHPQuebec organisation should be money driven.

First, let's go back a couple of years in the past. I remember going to the annual conference when it was held at the École Polytechnique de Montréal. At that time, about the same amount of important speakers were there, I remember first seeing Derick talking about PHP bananas or something. That year, I think the ticket cost me around 150$, I was paying student price.

Years after that, the conference moved to big hotels with somewhat bigger conference rooms, but most of the time with awful food. The ticket went up, but I didn't really care, for two reasons: I stopped going every year, and I could get discounted PHP developer rate for my contributions on PECL.

Last year, it was the beginning of the end. The conference moved to a very fancy hotel, which seems to have brought the ticket price even higher. It was now around 500$ if I recall correctly, with no rebate for core developers. I then have a very painful discussion with Yann over emails, asking him why that rebate was gone. He decided to offer me a rebate price after all, it was just not published on the website.

This year, it seems the ticket was still highly priced. I've decided to email Yann, Sylvain, or anyone in the organization, asking if they could provide a rebate for core dev. Beginning last December, those conversations were one way, I have never received any reply on any email I was sending.

Now here comes the tricky part. While waiting for their answers.. ticket price went up. I finally decided not to go. Talking about this with Philippe Gamache, he told me he would email the group, asking why they could not provide a core dev price this year. Their answer was clear: "We don't have any special rate this year".

And bam.. why would they refuse to answer me last December with those words? I could have decided to go for the pre-sale price. I seems that at the end, the only thing matters: the money.

Don't get me wrong on this one, I'm not trying to get a free lunch or something, and I am willing to pay a certain amount to go to the conference. This being said, I don't see why I would pay the full price. The reason having a discounted core dev price, is that you want them to be there. You want them to evangelize about PHP and you want them to pass the torch to younger souls.

The reason I want to be there, is to talk to certain people about PHP dev. I have no interest for the conferences, although I might go see a couple of them. Last year, I think I've spent more time outsite the conference rooms, talking with Marcus and watching David suffer while Rasmus was trying to hack his Mac.

It's too bad though, as it seems the PHPQuebec organization doesn't want to provide that ecosystem anymore. So go on, pay full price, don't invest in a better future for PHP, and have fun. I guess I won't be there this year.

DoubleMetaphone on PECL

nospam@example.com (Olivier Hill) — Thu, 08 Nov 2007 08:27:27 -0500

It's finally there... after quite some requests by external people, I've created a doublemetaphone module on PECL.

After some house-keeping, version 0.1.2 was born. It is still alpha, although it should not segfault or anything. The only reason going alpha is because I'm not sure the API should stay the same. After some feedbacks, I'll ramp-up beta and GA.

So, after a pecl install doublemetaphone, you can enjoy using the new double_metaphone() function.

If you have any comments, or found bugs, leave a comment here, or on the PECL bug system.

PECL GeoIP on Gentoo

nospam@example.com (Olivier Hill) — Mon, 20 Aug 2007 10:06:15 -0400

Even if the GeoIP bindings for PHP have been available for more than a year on PECL, I have never took the time to write an ebuild for it on Gentoo.

Until now, I was using the PEAR installer, which is great for installing PECL packages. The only downside is that those packages won't be seen when doing a world upgrade with Gentoo. You then have two choices:

Do an emerge world followed by a pecl upgrade-all
Install PECL packages via Portage

Since I'm lazy, I like the fact to upgrade everything at one place.

The hard part now is to write the ebuild, which is really simple considering the fact that it is a PECL package and that eclass are available.

One thing to another, this ebuild is now part of the testing overlay for Gentoo. If no bugs are found, it should be part of the normal tree shortly.

Kerberos kpropd on Gentoo

nospam@example.com (Olivier Hill) — Wed, 08 Aug 2007 14:58:48 -0400

Nothing fancy, but while configuring a Kerberos slave on a Gentoo server, an init script seems to be missing for starting kpropd.

While it seems some have difficulties writing init scripts, here is what I am using:


#!/sbin/runscript

daemon="MIT Kerberos 5 KPROPD"
exec="/usr/sbin/kpropd"

opts="start stop restart"

depend() {
        need net
}

start() {
        ebegin "Starting $daemon"
        start-stop-daemon --start --quiet --exec ${exec} -- -S 1>&2
        eend $? "Error starting $daemon"
}

stop() { 
        ebegin "Stopping $daemon"
        start-stop-daemon --stop --quiet --oknodo --exec ${exec} 1>&2
        eend $? "Error stopping $daemon"
}

restart() {
   svc_stop
   svc_start
}

DoubleMetaphone 0.1.1

nospam@example.com (Olivier Hill) — Fri, 29 Dec 2006 10:26:28 -0500

A couple of minutes after posting DoubleMetaphone 0.1.0, someone with a C99 compiler noticed that my mixed C/C++ code wasn't compiling.

The error? The use of the boolean true. So I rolled up 0.1.1, which is simply a four characters fix. You can download the new Double Metaphone for PHP version 0.1.1 here

For more information on the package itself, please consult the original post.

Double Metaphone

nospam@example.com (Olivier Hill) — Tue, 24 Oct 2006 10:17:59 -0400

Some time ago, I wrote a wrapper around the CPAN module Text-DoubleMetaphone by Maurice Aubrey for PHP. This allowed the use of the DoubleMetaphone algorithm with PHP. While the PHP core already has Metaphone, it lacks support for the second version, which handles latin languages a bit more correctly.

At the time, I did post an email to the PECL dev lists, but didn't receive any positive or negative answer. I thus never decided to publish it on PECL, thinking people weren't interested in the package - I was wrong. I have been contacted this week by email, asking if I could send the package to someone, if I still had the code. Well, lucky me, it had not the time to bitrot.

So here we go. If you need Double Metaphone inside PHP, grab doublemetaphone-0.1.0.tgz from here and follow the instructions on how to install PECL packages. You should be able to do something similar to this:

olivier@geeko-book $ pecl download

         http://www.olivierhill.ca/doublemetaphone-0.1.0.tgz

downloading doublemetaphone-0.1.0.tgz ...

Starting to download doublemetaphone-0.1.0.tgz (7,450 bytes)

.....done: 7,450 bytes



olivier@geeko-book $ sudo pecl install doublemetaphone-0.1.0.tgz

Now, don't forget to dl() the extension (bad...) or add it to your php.ini (good!)

If this works for you and you find it useful, you can always visit my PECL profile. It does have a wishlist URL if you have an urgent need to thank me.

geoip-0.2.0 released

nospam@example.com (Olivier Hill) — Tue, 22 Aug 2006 15:56:32 -0400

Finally, GeoIP is available on PECL. It's been a while since it all started, but now the first release is out the door.

First of all, GeoIP is a piece of software that can map an IP address or hostname to a geographic place. Although it cannot be 100% accurate (since your IP address is from your ISP which might not be next door), it gives you a rough idea in which city the user is coming from. With this data handy, you can then choose to present a local version of your website. Also, it can be really useful for demographic analysis where you calculate where your users are located based on log files. Anyway, what you do with that piece of software is up to you.

The good news though, is that it is now easily available from PHP. After long debates and licenses clashes (GPL being incompatible with the PHP License). The author of the GeoIP C Library kindly accepted to release the new version as LGPL. With this change, the GeoIP PECL module could be created, using the skeleton from SourceForge. Since that, many memory leaks were fixed, and documentation has been written. It should now be in good working conditions, let me know if you find bugs.

By the way, if someone builds a website using GeoIP to determine where the user is from, outputting this on a Yahoo! map using their API and fetching photos near where the user is from Flickr using geotags, leave a comment, I want to see this :)

stem-1.4.3 for PHP released!

nospam@example.com (Olivier Hill) — Mon, 23 Jan 2006 19:11:16 -0500

As of today, you can download a new release of the PHP stem interface to the Snowball API on PECL. While this extension has been written by Jay Smith, I have since joined him to help on further development.

If you do not know what a stemmer is, the article on Wikipedia is self explanatory. Basically, it allows a computer program to find a common root for different forms of the same word. While Dr. Porter did a great job creating stemmers for different languages and the Snowball API, it was not available directly from a PHP script.

Now that this limitation is gone, you might want to try using the stemmer to create an intelligent search engine for your website. If you want to give it a try, issue the following command on your favorite UNIX based machine: pecl install stem. Once the installation has completed, you might want to modify your php.ini to load the extension and then try the following example:

<?php
  print stem_english('cleaner') ."\n";
  print stem_french('épouses') ."\n";
?>

This would output clean and épous respectively. In some cases, the word outputed by the stemmer will not exist in a dictionary, but this is rarely a problem. In fact, you should only stem words to use them as keywords in some kind of database.

PECL on Gentoo

nospam@example.com (Olivier Hill) — Wed, 04 Jan 2006 13:04:51 -0500

If you try to install a PECL package without using Portage (thus using the PHP tool pecl), you might encounter an error like this snippet:


bender ~ # pecl install apc
downloading APC-3.0.8.tgz ...
[...]
autoconf: Undefined macros:
configure.in:63:AC_PROG_LIBTOOL
ERROR: `phpize' failed

The main problem lies with the use of automake v1.9.x. Since Gentoo comes with a bunch of different versions of the autotools, you can choose to use automake v1.8, which will result in a complete built.


bender ~ # WANT_AUTOMAKE="1.8" pecl install apc

As simple as it seems, it took me a while to fix it. Let me know if this helps.

PFE presentation

nospam@example.com (Olivier Hill) — Mon, 12 Dec 2005 10:04:59 -0500

The slides for my PFE (Projet de Fin d'Études) are available here.

It is in French, but if you are looking on how to integrate different technologies to build a full featured email server, you might have a look at some of the schematics. I will see if I can post the full report online shortly. The best would be to create an HOWTO, but it would be long to write.

My FOAF description

nospam@example.com (Olivier Hill) — Fri, 09 Sep 2005 23:48:42 -0400

After doing some readings on the semantic web and RDF, I came accross an emerging standard called FOAF.

What is FOAF? Actually, it is an acronym for Friend Of A Friend. What is its purpose? It let you describe yourself inside an RDF document with such infos as: your name, your email, your school and workplace.

But the juicy stuff comes right after. You can list your friends inside your FOAF description and link their FOAF to yours. As a result, you could build a giant tree of people knowing each others. Hypothetically, an RDF crawler could then parse the tree and permit queries such as: "Does Bob know Alice?" or "Show me Charlie's friends".

So go ahead, crawl my FOAF description and create yours today!

Tweaking Webalizer

nospam@example.com (Olivier Hill) — Thu, 11 Aug 2005 12:25:43 -0400

Looking at Webalizer's reports, I have just noticed that webpages returning a 302 do not get into the Total URLs listing.

I do not know if this is a bug or not, but a 302 is a Moved Temporarly status code. I do want to count those since it reflects how many people I am bouncing elsewhere. This simple patch does the trick:

--- webalizer.c 2002-04-16 18:11:31.000000000 -0400
+++ webalizer.c.new     2005-08-11 11:02:52.000000000 -0400
@@ -1080,7 +1080,7 @@
          
      /* URL/ident hash table (only if valid response code) */
      if ((log_rec.resp_code==RC_OK)||(log_rec.resp_code==RC_NOMOD)||
-         (log_rec.resp_code==RC_PARTIALCONTENT))
+         (log_rec.resp_code==RC_PARTIALCONTENT)||
+         (log_rec.resp_code==RC_MOVEDTEMP))
      {
         /* URL hash table */
         if (put_unode(log_rec.url,OBJ_REG,(u_long)1,

What would I become without having access to source code...

Playing with SPF

nospam@example.com (Olivier Hill) — Mon, 18 Jul 2005 22:11:05 -0400

Have you ever received spam that was coming from yourself or someone else from the same domain name? Well, I did. It seems to be a new trend that spam or viruses are coming from someone with almost the same email address you use.

But why is that? The answer is simple: whitelisting. When you filter spam coming to your inbox, it can be a common practice to whitelist your email address or the domain of your organisation so that it is never non-intentionally considered spam. This is good while working with Bayesian filters and the like.

So what is the solution to this common problem? Well, it is quite simple and it is called SPF, or Sender Policy Framework. Basically, what a sysadmin does is publishing a SPF record about which server should send mail for his domain.

Continue reading "Playing with SPF"