Stochastic Ideations

Link: http://doi2.acm.org/358453.358454

Back in March 1982, Peter Denning wrote (President's Letter, CACM 25, 3, pp 163-165) about "Electronic Junk". I stumbled across this article recently and was struck by both how much and how little has changed in the intervening 27 years, at least as regards electronic mail.

Professor Denning spoke of getting "5-10 pieces of regular junk mail, 15-25 regular letters, 5 pieces of campus mail, 5 reports or documents (not all technical), 5-10 incoming phone calls, 10-20 local electronic messages, and 10-20 external electronic messages." The ratios have changed for many of us: I get relatively little physical mail, far fewer phone calls, and far, far more email -- several thousand a day, before filtering, of which perhaps 50-150 per day actually get through to me. Probably 2/3 of those are newsletters or mailing lists that I read-and-delete (or just delete!) fairly quickly, leaving me with about 20-50 messages per day that require significant time and attention. Even if that only amounts to five minutes per message, I'm still spending 1.5-4 hours per day processing email. And of course, some of those messages represent multi-hour projects, but I think of those as being somehow external to the regular email stream just to maintain my sanity.

Professor Denning asserted the necessity of automatic filtering even before the advent of spam, as well as unfiltered "urgent, certified, and personal" mailboxes. He didn't predict dictionary attacks, believing that having "unlisted" mailboxes would be enough; we now know that essentially all mailboxes have to be filtered. He did discuss "threshold reception" (essentially a "bid for attention" scheme) that is a form of "electronic postage". He did not predict "proof of work" (e.g., hash cash) schemes, but it's interesting that he recognized the fundamental economic problem with email (cheaper to send than to receive) so early.

Today of course we filter essentially all email. We don't have "bid for attention" but we probably should. Many of us do use multiple mailboxes (or at least multiple addresses); I use a new address each time I sign up at a web site so that I can turn the address off if it starts getting spam (which also turns out to make it easier to recognize legitimate mail!). Sender authentication such as DKIM will move ideas such as "certified" mail at least into the realm of possibility, although more work is of course needed.

He didn't get everything right. He proposes "importance numbers", essentially a way for the sender to assert urgency, without realizing that malicious senders would lie. He talked about individuals having multiple mailboxes without predicting the problem of managing more than one or two. He mentions "restricted access" mailboxes without mentioning the necessity for authentication; in fact, he seems to have failed to anticipate bad actors at all (spammers, phishers, etc.) -- although he did acknowledge junk mail. I think he probably thought of junk email, but not in the quantities that we see today. Interestingly, he talked about certifying message quality, but he didn't mention sender quality (what we would today call accreditation and reputation).

Despite these gaps, he understood the fundamental problem of managing information overload and that email was going to make things worse in the short run. There are some interesting techniques for that, but that's another topic.

For a contemporary reaction to his topic, see the June 1982 ACM Forum (CACM 25, 6, pp 398-400).

I've got several machines on or under my desk at home: my FreeBSD server, a mac mini, and a laptop, but only one keyboard and monitor. No problem, just use a KVM, right? Well, I'm on my fourth (or is it fifth?) KVM switch and I'm still trying to find one that will do what I want.

The big problem is that I have a "media keyboard", which I actually use (at the moment, I'm using the Logitech Wave keyboard). There's a great Mac app called ControllerMate that lets me do arbitrary actions on USB devices, so for example I can control the lights in my office from the keyboard using an Insteon connection. But I digress.

Most KVM switches want to create a "virtual keyboard" so that they can (for example) make the keyboard appear to remain connected while you're switched to a different machine. They do the same with the video connection so that your computer doesn't think it's been beheaded when you switch to a different machine. Here's the problem: the media keys on the keyboard appear as a separate USB device (usually a mouse with a gazillion buttons) , but the KVM switches map them to an internal convention, thereby disabling many of the keys. I'm sure the manufacturers think they're doing me a favor, but they are wrong.

The best option I've found so far is the TRENDnet TK-407K, which has no such smarts at all. It's just a raw path -- no virtual keyboard, no virtual mouse, no virtual monitor. In other words, it just stays out of my way. Incidently, it's also probably the cheapest KVM switch I've found. Because there are minimal active electronics it can be pretty high res (I'm running 1920x1200; it claims to work up to 2048x1536).

But it isn't perfect. The main problem with it is that when you switch machines you have to wait for the selected machine to detect the "new" hardware. Fortunately this is pretty quick on MacOS and FreeBSD, and I don't have any Windows boxes any more (I use Parallels when I must). But what's worse is that sometimes the USB doesn't seem to sync up at all (the light just keeps flashing and the keyboard and mouse are dead). Usually I can fix this by switching to another KVM port and then back again, or by unplugging the USB port from the host and plugging it back in again (easy for the laptop, hard for everything else). Sometimes I have to do this several times. I have no solution for this, and it really gets in the way of getting work done.

One other minor glitch: two ports come out the back, plus one on each side, and the keyboard and monitor connectors are on the front. The buttons are on top; there are no keyboard escapes, so you need convenient access to the unit. It's small out of the box (about 17x7x3 cm), but then you have to plug in the (proprietary, but not usuriously expensive) cables. Once you account for the bend radius in the cables, plan on dedicating about 40x30x3 cm on your (physical) desktop to the KVM. Why couldn't the cables have just come out the back?

I would have stuck with my old Raritan SwitchMan SW4-USB if it had just treated the media portion of the keyboard as a separate device that shouldn't be mapped. The Raritan was really a good KVM switch in every other respect, with a separately switched USB hub so I could have my printer connected to one machine while the keyboard and mouse were on another. Ah well.

Link: http://www.ceas.cc

The Conference on Email and Anti-Spam (CEAS) is coming up August 21-22. I've gone to this conference since the beginning, and so far as I know it's the only conference on the topic that is practical without being arm-wavy. Good stuff if you're in the business, and there's a lot of interesting work there. IIRC that's the first time I heard definitive research showing that CAPTCHAs have essentially been defeated (using algorithms, not presenting them to other humans), and that was at least three years ago. The rest of the world is just figuring that out.

Another good conference is the MIT Spam Conference. However, I don't find it as rigorous as CEAS. Unfortunately I managed to miss the MIT conference this year. Although I'm on leave and hence don't have a travel budget, I probably would have tried to go.