Mark Watson's opinions on Java, Ruby, Lisp, AI, and the Semantic Web

As a Linux user since the early 1990s (and a longtime OS X user), it was easy for me to get in the "./configure; make; sudo make install" habit, but I don't think that this is such a good idea for two reasons:

• Security: have you really read the source code to see what might be executed during "sudo make install"? I am constantly installing Ruby gems, infrastructure software, etc. and I often read code as an educational experience, but not for security. It is best to not run other peoples code as root.
• It is much easier for me to rebuild systems from backups when I "./configure --prefix=/home/mark/bin" (or wherever, but in my home directory).

I used to like to keep my home directory fairly small so backups take up less space but now costs of external disks, remote storage like S3, etc. are so small, that it makes more sense to have my home directory to be more self contained.

I also like to develop customer projects under a single master directory. It is nice to have everything in one place: my application code, nginx, PostgreSQL (with data), Ruby, gems, Java, Tomcat, Sesame, Erlang, CouchDB, etc. - whatever a project requires to run. A top level shell script can set up the environment for each different project. This also makes cloning a customer's system to one of their alternative servers just a quick rsync away...