Bifurcation of Knowledge

My friend Andrea Dainese (of the Route Reflector Labs fame) sent me this observation:

Because of lack of fundamental skills, I see two groups forming: junior guys with low salary (the bigger group), and a few experts (hopefully with higher salary). The middle group is disappearing. Intermediate-level engineers are either moving to the entry level (because the complexity is increasing and they are not keeping up with it) or to the upper level.

I call this phenomenon bifurcation of knowledge (I’m positive it has a formal name – would appreciate a comment with a set of pointers), and it’s a direct result of commoditization and the changing shape of the learning curve.

Not surprisingly, I wrote about it before. You might want to read:

It’s also not limited to networking. When I was at university we spent half a year on developing a compiler (and probably more than that on various finite-state automata and parsers). While most everyone found it a nuisance best forgotten, it gave me enough knowledge to build my own PHP-like interpreter I needed to replace dBase and Clipper with something resembling a programming language. How many programmers even know what a compiler does these days?

Too contrived? How about “do you know how bolts are made” or “could you make one if needed”? BTW, I could – I even have the tools to do it ;)

Is bifurcation of knowledge bad? Not necessarily. As we cannot cram infinite knowledge into our brains, or spend infinite time acquiring it, levels of abstraction and ignorance of details are the only way forward.

However, there’s always the pesky Law of Leaky Abstractions. There’s not much that can go wrong with a bolt made in standard dimensions from materials of known quality… and plenty of things that can go wrong with any black box a startup hacked together, a major vendor acquired, an a marketing team spray-painted with unicorn dust.

Also, people working with physical objects (like engineers using bolts to hold together a bridge) tend to be aware of laws of physics and limitations of objects they’re using, while you’ll find copious amounts of firm believers in Clarke’s Third Law and the unbreakable magic of whatever gadgets they’re using in IT (see also: long-distance vMotion).

The real difference between engineers (in the traditional sense of the word, not as a slap-on job title you get instead of a raise) and some IT practitioners is the ignorance of the fundamentals. While you don’t have to know how 1GE encodes bits on the wire, it does help to understand a bit of history, the differences between cut-through and store-and-forward switching, and why full-duplex and half-duplex matter (although they should not since the days of first L2 switches… but please don’t get me started).


  1. This comment has been removed by a blog administrator.
    1. I don't miss the days of being CLI jockey ;) Looks like our paths are parting... too bad our acquaintance was always one-sided.
    2. Don't be silly, it's not about CLI.
  2. I think the main driver is money. I used to work for a company where twe had alway 5 CCIE employed. Never 6+. Every year new employee used to pass CCIE exam. But 6 was too much and once we had the 6th CCIE, one of them had to leave the company. At that time we always had A-, P-, and E-level guys. So there were always people at the middle-level dreaming about being CCIE.

    What I want to say: I do observe this bi-polar World now. The A-level guys believe in google & stackoverflow powered engineering.
    These guys are cheaper and solve most of the technical problems (because most regular problems are solvable through power of google as most problems are not very unique).
    And only few well paid E-level guys are employed as 'master-minds'. Like we (in the past) needed only 5 CCIE and were not willing to pay more for E-level staff.

    I see another fenomena - the E-level staff is not willing to share their knowledge with the youngest generation...
    1. >But 6 was too much and once we had the 6th CCIE, one of them had to leave the company.
      >I see another fenomena - the E-level staff is not willing to share their knowledge with the youngest generation...

      one can only wonder why, in such a wonderful company.

      CCIE by all means is not a sign of being on the top. It's like a first black belt in karate - just a beginning of the real journey
    2. I was writing about times around 2000 year. Brain dumps were not available at least in my country a that time. People had at least 7-8 years of experience in nation wide network deployments activities.
      Yes, CCIE now is not as important as in XX century;)
  3. It's a typical hourglass effect in a mature market.
  4. I think that nowadays there is something like a fundamentalism with SDN (or "SD..."). Pretty much everything today in networking is pretended to be solved with a solution based on this magic term. I compare this to what happens with "democracy" in most of the occidental societies.

    First, the term is vague, imprecise. A lot of engineers think they know what´s the meaning (a known unknown) but just a few really know what´s the accurate definition. Some vendors use this to sell their products and to say that it solves most if not all the problems/inconveniences that traditional networks have. It happens the same to democracy: the term is vague and almost anyone think they know what it really is. However, the accurate definition is known by just a few. In a similar way, some politicians use a lot the word democracy to say that this magic word would solve most if not all the problems of the citizens.

    Second, sometimes the vendors do not provide documentation in depth describing their SD.. solutions. Therefore, your option is to trust and accept or untrust and reject. I think that politicians sometimes do something similar, avoiding commitment and relying on democracy and promises.

    As a consequence of that, I think that some engineers believe in the SD.. like a gospel without really understanding the solutions, or at least asking and pushing the vendors to provide more documentation than just a ppt and a demo. I see some lack of critical thinking in networking nowadays.
  5. Totally agree with you guys. The funny thing is that I guess in the end money will be spread just differently in terms of expertise. Fewer low-profile engineers as automation is gonna do their work but more high-profile engineers to understand, manage, design and troubleshoot (all supported by the automation of course) the added complexity of the automation itself.
  6. back in the day on mainframe networks, job roles were voice, network hardware and network software. the software guys did the automation and programming.
    just another case of whats is old is new
Add comment