I'm stuck for interesting blog topics.

[katster]

#1: Cal lost to USC tonight. Still, it was a decent season for my Boys in Blue, and we've got the NCAAs ahead. While it would have been nice to win the Pac-10 tourney, it's not the end of the world.

#2: It's raining and windy in Berkeley tonight. I like rain and wind. And I got to be out in it for a bit while I was waiting for my bus. So no urge to go out and play in it when I should be sleeping.

#3: Lies on Applications prompt Haas to reject 5 percent of applicants. Haas, for those of you who aren't in the know, is the UC Berkeley business school. They ran a background check on all their applications and found that five percent of them had lied on applications and/or resumes. They're doing this in the wake of last year's corporate scandals. Another thing they're planning to do is talk about ethics in all their classes. I applaud the business school for doing both of these.

#4: Speaking of b-school, I continue to maintain that I really don't like this "SIMS is a professional school" mindset, and if I'd known about it beforehand, I might have looked for something else. But I'm here for the duration, so I might as well make the best of it. I'm going to try to do fun and interesting things in the next year, and maybe try to take the courses Buckland would suggest to get him to write a letter than my degree is "functionally equivalent" to an MLIS. I think I'd prefer being a librarian more than a glorified b-school grad.

#5: I should really try to get life@SIMS running this weekend. And do homework.

#6: There were other things I meant to talk about, namely the resolution (for the most part) of my housing situation, but the short of it is that Luns didn't say what I thought he said. And because of this piece of news, it means I can tolerate Jamila a whole hell of a lot more.

#7: 206 midterm Tuesday. Whee. On the other hand, I talked with Prof. Chuang right after class, and he said that a 12 and a 13 on my quizzes was 'quite good.' I remain dubious, but he's the prof...so...and the midterm covers networking, which is right up my alley. In fact, here's a midterm question from last year.

"NASA is planning to build a colony on the red planet, Mars. They want to build a communications network for the colony, and link it to the Internet on Earth. In addition to the two terrestrial networks (one on each planet) there will be one set of satellites in Mars' orbit and one set of satellites in Earth's orbit. The two sets of satellites will act as routers for all network traffic between Earth and Mars. The satellites are positioned such that, at any given time, exactly one Mars-orbiting satellite will be able to communicate with exactly one Earth-orbiting satellite.

"As a summer intern at NASA, you have been asked to evaluate the applicability of the TCP/IP protocol suite for this network.

"(a) Discuss the addressing function of the Internet Protocol (IP) in the interplanetary context. Are the IPv4 or IPv6 addressing schemes adequate for this purpose? If not, what changes would you recommend for version 7 of IP to handle interplanetary networking?

"(b) Discuss the packet switching functions of IP in the interplanetary context. In what ways might the satellite routers differ from the traditional terrestrial routers in their implementation of packet switching?

"(c) What are some potential problems of data transfer using TCP, given the fact that a data packet takes 3 minutes to travel from Earth to Mars (moving at the speed of light)?

"(d) How would you overcome these problems, with or without making modifications to TCP? "

Have I mentioned that question completely and utterly kicks ass?

#8: A possible reason girls don't go into technology? It seems to ring sorta true for me, except for the fact that I *hate* collaborating, and would much rather do other things with my time. But interesting thoughts nonetheless. Link yoinked from Doc Searls' blog, which, btw, is worth the general read. You can find his LJ rss feed at docsearls, btw.

#9: I need to get better about going to bed closer to my weekday bedtime on weekend nights. That said, g'night!

Comments

[miggy.livejournal.com]

Haas, for those of you who aren't in the know, is the UC Berkeley business school. They ran a background check on all their applications and found that five percent of them had lied on applications and/or resumes. They're doing this in the wake of last year's corporate scandals. Another thing they're planning to do is talk about ethics in all their classes. I applaud the business school for doing both of these.

That is awesome. I love it.

[smcwhort.livejournal.com]

Oh, #7 is just too delicious to pass up ...

(a) IPv4 and IPv6 addressing are both more than adequate to the task. The only difference between the Mars-network and the Earth-network are that there are some very long distance hops. The general packet routing scheme that works on Earth, with packets finding their way to the right routers, will work in general.

(b) Satellite routers will have to deal with hertofore-unknown-to-TCP/IP latency. They'll have to assume and prevent worst case scenarios. For example, packets *will* be multiply retransmitted before any indication of failed reception comes back from the opposite planet, since we can crank bandwidth up via parallel transmission streams as much as we care to. To ensure that TCP/IP TTL will work out correctly, satellite routers will only decrement the TTL field by one second no matter how long packets take to get from planet to planet.

(c) TCP/IP itself *can*, in theory, stay more or less as it is, but with real-time timeouts extended to ridiculous extents. This implies that data in retransmission windows will have to stick around for many many minutes at a time, possibly causing TCP/IP implementations to have to buffer many megabytes (or perhaps even gigabytes of data, depending on bandwidth) for extended time periods. This may be solved by low-level redesign of the protocol which takes into account the revised retransmission scheme implemented by routers on huge-latency network hops.

The real problem in this scheme is that virtually no protocol that is based on TCP/IP is designed with such ridiculous latencies in mind. 3 minutes is a *best-case* scenario; the worst case scenarios are easily over 20 minutes of latency that'll occur when Earth and Mars are on opposite sides of the sun. Imagine HTTP, with its chatty back-and-forth fetch-this-and-that from the server conversation scheme, or POP3 with its equally chatty mail-fetching scheme, or ... virtually all TCP-based protocols.

Almost all such protocols will have to be redesigned to send much much larger chunks of data by default. Protocols that encapsulate large chunks of data will have to be designed. Most interaction with other-planet data will take place through large caches on the local side of the link. A UUCP-style strategy for transmitting data may re-surface as the best way of transferring data. UUCP is somewhat designed to manage large-latency links, as it used to use the local telephone dial-up networks to trade data, which imposed equivalently large latencies. Usenet-style bulk data transfer from server to server across the hop will probably come to dominate the interplanetary links.

(d) Given the discussion in the answer to (c), I envision TCP/IP being an *extremely* tiny component of cross-planetary traffic. Perhaps we'll see schemes similar to TCP/IP over UUCP-style network links. Perhaps there will be a new MIME-type: network-packet/ip ... (talk about putting the cart before the horse!) to manage cross-planet TCP links.

being silly

[siliconshaman.livejournal.com]

Wot, no subspace radio ? No quantum tunnelling transmissions with zero transit times ?

Tck, the future ain't wot it used to be !