Development

The public LikeOrHate API is getting there: we finally released the first full featured beta version (and you can read all about it). We're pretty certain any changes will be minor until the 1.0 release is made, which should happen within the next month.
With the API it will be easy to import data from #LikeOrHate# to any other site or even application. It's a REST API, so it's very simple, but we are also developing libraries for a bunch of languages (PHP, Python, Java and Javascript for starters).

We are testing the REST services and writing the APIs for other languages. This is an interesting exercise, albeit usually boring: writing the exact same code in many different languages. We're starting with PHP, Python, Javascript and Java, which should cover a good deal of the needs of our developer-users. Other languages will come, and we'll gladly accept your own translations.
#PHP# and #Python# are remarkably similar when it comes to small programs. It's almost possible to translate one to the other with nothing but a few regular expressions. The only big difference is how to make HTTP requests. Java will be written by someone else (sometimes I'm lucky, hehe). Then there was JavaScript.
Our #Javascript# code that makes these nices boxes when you hover your mouse on the LikeOrHate links changed a lot during development, and though it's quite clean and easy to work with, it doesn't have what could be separated as a public API for accessing LikeOrHate's REST services. So I set out to write a new one, thinking it you be easy to port from the PHP/Python code. Indeed it was, except for one thing.
You can't make synchronous cross domain request in JS. Ridiculous. You can either make it synchronous using AJAX, but then you hit the cross domain problem, or you can make it asynchronous with <script> tags.
As a result, all JS APIs work by making requests and getting callbacks. And so will ours, to my great dismay.

We're proud to announce a sandbox for developers, which is available at http://sandbox.likeorhate.com. This way you can develop your application and test it, without messing with the main database -- the sandbox is reset periodically.
More information about the sandbox is available at the help page.

After the crash I had to download bunches of software, among them #VMWare#. I love their software. I use it often for development -- it's a nice thing to keep a project in a virtual machine and know that I can mess the environment as much as I need without fearing the disruption of my daily PC use.
Except downloading it is a pain. The package now weights 500MB (ouch?) and you need to login to download. No wget for you. Why that? You still need a key, so why block direct downloads? So unpractical.

A couple days ago my computer was suddenly very slow. There were no apparent errors I thought that Eclipse was hogging the PC again. I shut it down, still slow. Fine, just reboot.
Tons of errors on my primary HD. It still could be accessed, but barely.
Ever since my first disk crash (caused, too many years ago, by doublespace; if you know what doublespace is, you've been using computers for too long), I started to backup more or less regularly. So I lost almost nothing important, have everything I've made up to a few weeks back and thanks to the cloud not even one email (except for a couple accounts which I don't keep emails on the server). SVN is also very helpful in these events. All LikeOrHate code anda data is completely fine -- even the bleeding edge development stuff.
It was less traumatic than expected, though I still am reinstalling everything. It killed my linux partition, which was the same (moved through many HDs) since 1996. Perhaps it was time to clean it up. Unfortunately, some stuff I didn't backup (media files) may be gone for ever. I could take the disk to have the data back, but it's not worth the price they charge. Perhaps the Gods Of Computers will let me get some of those gigabytes back before I convert the old HD into #frisbee#s.
And perhaps that mirrored RAID that I used to consider only for servers is not an overkill for my desktop either.

Here are some tips for those of you running #CentOS# who would like to run #SphinxSearch#, an awesome full-text search engine, in SphinxSE mode (that is, inside #MySQL#):

1. Install MySQL 5.1. It supports plugins, so you don't need to recompile the entire thing to have SphinxSE. Since 5.1 is not yet the default in Centos, you can get it from the Remi repository. Read here to setup remi in yum.
   
2. Install/upgrade mysql: "yum --enablerepo=remi install mysql-5.1.34 mysql-devel-5.1.34"
3. I'm following the idea from this page. Get the mysql 5.1 sources that match the version you installed above. Unpack. Copy the sphinx-x.x.x/mysqlse directory to mysql-5.1.xx/storage/sphinx.
   
4. In the mysql source directory, run 'sh BUILD/autorun.sh; ./configure; make'.
   
5. Have a coffe.
6. Run 'cp storage/sphinx/.libs/ha_sphinx.* /usr/lib64/mysql/plugin'. Your plugin directory may not be this one, check it in my.cnf
7. From mysql, as a superuser, run "INSTALL PLUGIN sphinx SONAME 'ha_sphinx.so';"

If you are primarily a developer, like me, chances are your design skills suck. They may be functional, they may even be nice every now and them, but people look at your sites and they say, "Yeah, the design is not good," or "the idea rocks, the layout sucks," or my personal favorite: "I like it, but it looks too square."
Since you've studied algorithms, half a dozen programming languages, the two thousand, three hundred and eighty six web standards (and counting), databases, system administration, and probably business management and accounting (if you are opening a #startup#), why not study design? Oh, yeah, you've googled for "design howto" and nothing useful came up. So here are a couple books to help.

• "#The non-designer's design book#", by #Robin Williams (writer)#. This is a wonderful book, clear, concise, simple. It's not about web design (but it talks about it a little bit) and reading it will change the way you see any material with words on it. Read it.
  
• "#Don't make me think#", by #Steve Krug#. This is almost a classic about web usability. It is much more practical to web than "non-designer's", but you can enjoy it much more if you have read "non-designer's" first.

Good reading.

So you need to backup your #MySQL# database. Easy. Ah, you don't want to bring your site down while you're doing the backup? Hard.
It's amazing that a product as widely used as MySQL has been so lacking in a fundamental feature such as backups (well, perhaps people really don't backup). It's true that version 6 will have a BACKUP command, but you don't run alpha releases in production servers.
The most common solutions for hot backups in MySQL are using replication, LVM and mysqldump. Now there's a new tool called mydumper. It's quite new and it doesn't support MyISAM yet, but it's worth checking out. It's quite faster than mysqldump: http://dammit.lt/2009/02/03/mydumper/.