On 24 September 1993, the IETF published RFC 1519, designating Classless Interdomain Routing (CIDR) and variable length subnet masks as the standard. That particular document is obsoleted by later RFCs, but it’s still a milestone.
Before then, IP addresses were allocated by “classes.” Class A, B, and C addresses were the norm. I’m not going to explain classful addressing, because it’s long obsolete and, on the current Internet, stupid.
What I am going to do is go on a mini-tirade about classful addressing. Because there’s a lot of people out there still teaching classful addressing to newcomers. And then these poor newcomers hit the field, and people like me have to spend our time unteaching them what they so painfully learned.
I fully understand it takes a few years to disseminate knowledge. But textbooks are still being published that claim classful routing is the standard. This is an appalling disservice to the profession.
Yes, CIDR looks hard. But if a new network admin can’t handle CIDR and VLSM, they shouldn’t be administering networks. That’s perhaps the easiest math they’ll need to handle in their career. And the Internet is full of cheat sheets for people who don’t want to bother to do the math.
On this, the 25th anniversary of Classless Inter-Domain Routing, I hereby declare 24 September 1993 CIDRDay, dedicated to stamping out classful addressing. A whole variety of celebrations are appropriate.
First, of course: cider! Cider is obligatory on CIDRDay.
Second, whenever someone who should know better says “Class C,” “Class B,” or “Class A” address? Explain to them the error of their ways, with the minimum amount of force needed to make sure that they never say it again.
If you know someone who’s still teaching that garbage? Yell at them until they promise to stop. If yelling doesn’t work, escalate.
Because frankly, I’m tired of reeducating innocent newcomers who should have been better served by their instructors.
Fully Concur
*applauds, then hoists a mug of his favorite apple brew*
_However_, as a historical note, would you endorse mentioning that previous scheme *as a way of (partly) explaining current IPv4 allocation*? (I’m thinking of A Certain Large Automotive Manufacturer, who, because they got in on the gold-rush early, has an entire “Class A” block allocated to them, etc.)
Once you get into history, sure. Classful addressing is a valid part of history.
Tell me you’re issuing me a class C and you’re gonna get a face full of knuckles.
“Because early routers and systems and system administrator brains were not very powerful, once upon a time you could only have a /8 IPv4 subnet mask. Networks using a /8 mask were simply called… networks. Later on (RFC791) you could use three different subnet masks that each ended on byte boundaries, and IPv4 networks that used the three specific bit positions /8, /16 and /24 were called, respectively, Class A, B or C.
Then things got really confused because Classes D and E meant something else entirely. Thankfully this is mostly[1] a historical footnote now, and subnet masks can (theoretically) be any length.”
Class D and Class E are still sort-of-valid things to mention, in certain – very unfortunate – circumstances. My condolences if you ever need to care about Class E, but the Class D bit pattern is baked into a bunch of silicon out there, notably NICs and switch fabrics. Thankfully we can talk about them using VLSM notation, too.