Diary for herman

Older diary entries for herman (starting at number 18):

7 Nov 2000  »

Pictures of the Internet

Since I'm covering BGP routing and the Internet backbone, some pictures are in order.

3 Nov 2000  »


I've graded the second exam and fourth quiz. See the grades page to see the scores. If you're curious about the solutions to the second exam, look here.

19 Oct 2000  »

Exam 2

It's not graded. I'm not sure when I'll be able to grade it, given other time demands. I did post an update to the reading assignments for the next week.

16 Oct 2000  »

cleaning up sliding windows

I skipped over several important topics in my lectures on link protocols. These are explained in the textbook, but it will be helpful to mention them explicitly.

  1. variable frame size. In the examples I've presented for calculation of window size, effective bandwidth, and so on, there is an assumption that frames are all the same size (only the ACK is smaller). In fact, many protocols allow for varying frame sizes.
  2. full-duplex, half-duplex, and assymetric bandwidth situations. There is some relation between how window protocols work, the best design for parameters of these protocols, and how bandwidth is allocated in a channel between two endpoints.
  3. NACK (Negative ACK, sometimes also written as NAK). This can be a complement to ACKs, an alternative technique to using ACK packets, or convey extra information from receiver to sender.
  4. two-way sending. In practice, on many channels, both endpoints of a channel have data to send. In these circumstances, the sliding window is used simultaneously by both sides, each endpoint acting both as a sender and a receiver.
  5. piggybacking. This term mainly arises in the context of two-way sending. Why send a separate ACK packet when you also have a data frame to send? Why not combine both into one packet? That's the essence of piggybacking.
None of these points are essential to understanding how the basic window protocols work, but they are essential to getting ``the big picture'' of how things work out in practice.

15 Oct 2000  »

course notes

I'm having some difficulty in preparing and posting course notes. Finally I've posted the notes for Thursday's class (several days late) because I think there is some error in one of the calculations, but I couldn't remember exactly where. The next notes on Ethernet will require some figure drawing that may be too time-consuming.

10 Oct 2000  »

Second Exam

It's not too early to start thinking about the second exam. So I've posted an article and made pointers to previous exams. Now I can concentrate on preparing the remaining three lectures before the exams. This time I hope to throw in a little about wireless techniques not covered in the textbook.

Also, I need to think about what lectures (if any) I should prepare while I will be away during the week of 22-27 October.

9 Oct 2000  »

Busy Week

As the book states, the sliding window protocol is perhaps the best-known algorithm in computer networking. It illustrates many themes: efficiency, error recovery, flow control, and the implementation of a reliable service from unreliable hardware. This week I'll very quickly cover this material, leading up to the next exam. First, I'll just quickly describe the algorithms in a high-level fashion. (Actually, I won't have time to do it in detail --- that's what the textbook does.) The second lecture will concentrate on analysis and calculation.

4 Oct 2000  »

Quiz #3

I graded the quiz. The questions were difficult, mainly because I asked questions about material in the textbook but not discussed in a lecture. The material about protocol stacks, multimedia channels, and related terminology is important --- but just as easily learned from the text as by lectures.

Actually, it may be good to have low quiz scores, if this indicates students are concentrating now on the project (which is far more important for the final grade) rather than on reading.

Trip in October

I'm scheduled to be away, attending a research seminar, during the week of October 22-27. This means I will have to miss two classes. The project next week is to tape these two classes ahead of time, meaning that I'll need to get busy with preparing lecture material. Also, October 19 is the date of the second exam, so that will also mean some extra preparation for the coming week.

27 Sep 2000  »

Proxies, HTTP, and HTML

Basically, the problem for someone learning how web servers, TCP, HTTP, HTML, and related things fit together is: there is no proper place to start! I posted two articles on some tools (an echo server and a tcp proxy) that I've found useful to learn how netscape and webservers really work. But in order to make sense of these tools, one has to know something about networking (ip addresses etc), unix shells, web page caches, and hundreds of other small details. I haven't even got to any of the official references on HTTP and HTML (maybe w3c is a place to start, but it's intimidating). I'm hoping that students will somehow find references online, in books, or from each other as well as asking me questions.

25 Sep 2000  »

Grading of First Exam

I graded the first exam and posted the results. Generally, I'm disappointed because the grades are too high, which could mean

  1. the exam was too easy
  2. students came into the course knowing it all
  3. students spent too much time preparing
  4. the professor spent too much time preparing students
I tend to think number 3 is the case, meaning that students are putting too much emphasis on exams --- since 50 percent of the grade is for projects and so far very few students have even posted diaries about the plans, experiments, and progress on their projects.

Part II

This week really starts into the second part of the course. After introducing network measurement fundamentals (covered in Chapter 1 of Peterson/Davie), I'll work problems during one or two lectures, because there is little discussion in the textbook about how to solve the excellent exercises at the end of the Chapter 1.

9 older entries...

[ Home | Help | Articles | Account | People | Projects ]