![]() Instead of expecting humans to adapt to an interface slapped together by engineers, engineers should strive to create a system that is designed for humans in their natural form.ĭr. When engineers just love this stuff design something, we wind up with a system that asks an innocent person attempting to install email, "Do you want POP or IMAP?" One common comment engineers make is that we need "more user training". Though, I've learned that some people like to configure things, so I usually design in knobs for them to play with, and perhaps improve things, but any setting of the knobs will still work correctly. Things will just work, and be self-managing. The stuff I design, if I'm successful, nobody will ever notice. ![]() Then I'd have her friends calling me up and saying "when I try to configure a printer I get this error message." I had to explain to my daughter that, yeah, although I'm considered an expert at this esoteric thing called network protocol design, I wasn't likely to be able to help her friends at anything practical.īut the world would be a better place if more engineers, like me, hated technology. When my daughter was in high school she told her friends proudly that her mother "had a PhD from MIT and knew all about computer science". ![]() ![]() Is your love for technology what inspires your work?Īctually, I hate technology! There are some people who love gadgets, love using them, love taking them apart and tinkering with the insides. You have transformed the technology industry several times with the spanning tree that transformed Ethernet, with link state routing, and with TRILL. I believe that understanding design tradeoffs and alternatives will be far more beneficial than memorizing the details of one protocol suite, and will enable the students to think critically, and will increase their ability to design new things. Instead, I believe students should be taught the concepts, such as "how would one acquire a network address without being configured", and then study a bunch of different approaches that could exist, or have existed in other protocols. Rather than approaching it as a science, students are merely told to memorize the specifications of the existing deployed devices. I get frustrated with how networking is taught in many places. What changes do you think need to be made to the way engineers and computer scientists are taught? Many network protocol designers learned the field from your books, and keep a copy on their bookshelves. And, try to think about and understand various approaches, and the tradeoffs between various choices. This is a combination of "what customers are asking for", "what customers don't even know they want yet" and "what can be solved with something simple to understand and manage". Start out with finding the right problem to solve. What advice would you give to fellow engineers? ![]() TRILL will fix these problems, but do it in a way that is completely compatible with everything. While spanning tree-based Ethernet was popular because of its simplicity and ability to self-configure, it didn't make optimal use of bandwidth, and was, for subtle geeky reasons, somewhat fragile. Ironically, one of the things I've been working on these days is replacing the spanning tree with a new technology, being standardized in IETF, called "TRILL" (Transparent Interconnection of Lots of Links). ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |