I know I shouldn't feed the troll but I can't help it. I can tell you've never used both in more than a cursory way. They have very similar functionality now. Except RPM and yum is actually better with multi-architecture (which is very common with 64bit/32bit mixed on a 64bit system) than apt/dpkg.

And yum is every bit as usable as apt. So I'd say actually yum/rpm has the upperhand until everything goes single architecture again and the migration to 64bit is over. Or if someone fixes apt/dpkg.

On the other hand, for a desktop, the end user should normally never see either. They are likely to see synaptic or some front end.

75$ in 1897 would be a large sum as of recently:

What cost $75 in 1897 would cost $1846.03 in 2007.

According to one inflation calculator. But the linked article as a man in 1989 cursing and that 75$ isn't as interesting:

What cost $75 in 1989 would cost $123.93 in 2007.

As with all recommended books- make sure you read critically:

System and Practice of Network Administration by Limoncelli and Hogan. Not a how to book a why to book. It should be required reading for everyone in IT.
How to Win Friends and Influence People by Dale Carnegie. Also should be required reading for everyone in IT.

If you are a router jockey:
Routing TCP/IP vol 1, by Doyle, covers the IGPs.
Internet Routing Architectures by Sam Halabi

And the new world:
MPLS and VPN Architectures (probably vol. 1 and 2 if you have to do Service Provider or VRF) by Pepelnjak

If you are a sysadmin- you should read every shred of manufacturer's documentation on their website especially the login required. But if you can't always read the installation and configuration guides.

If you are a software dev guy:
Mythical Man Month- Fred Brooks
Peopleware- DeMarco and Lister

It will teach you about the why and how of managing the development cycle. Of course the algorithms and tools, and languages books are important, but so is understanding the development cycle and how the rest of the business sees it.

I wish I had a good intro to business text for the slot to recommend to all the types.

There is Out of Crisis by Deming for managers. I could probably come up with more...

??

1- NFS performance is amazing. It isn't the protocol you have performance problems with it is the transport (layer 1 or layer 2). The protocol in a transport might make a couple % points difference, and that even rarely.

The transport is where it is at. Comparing gigabit with FC is a losing battle for NFS, but compare 10G with FC (even 8G FC) and you have NFS at the top of the performance heap right now for mass storage, only iSCSI is in the same ballpark- but it is also on... 10G ethernet. iSCSI also cannot do simultaneous reads/writes like POSIX compliant NFS can. Direct attach is miserable because you invest loads in disk and can only use it on one server. What if you want to share that data around? Replicate? Islands of storage?

2) use automounter. Seriously, this hasn't been a problem for 5-10 years. Automounter, hostnames, don't use IP addresses (better if you can reverse the addresses).

You obviously haven't maintained NFS either recently or in a large environment.

NFSv4 does things your post doesn't even mention (security and ACL improvements, some performance in some cases).