From: tcmay@netcom.com (Timothy C. May) My point? Much wheel-spinning. Like trying to read Amanda's "X11" GIF, and then wondering if my Netcom disk quota was being sucked up by a hidden file somewhere! Or jumping through hoops to download a PGP-encrypted note to my home machine, decrypting it, only to find a "Like, wow, this PGP sure is neat! Like, rock on, dude!" message awaiting me! [...] There's got to be a better way. Tim's rant is one of the best illustrations of the effects of transaction costs I've seen recently. Tim's story perfectly illustrates the reason why the computer software industry doesn't move faster. TYLISUM -- Ten Years Later I Still Use Microsoft. The costs here are the of transaction of switching software systems. In order to understand exactly what the transaction cost is, we posit two worlds with respect to, say, email handling. World 1: The status quo. Adequate capability. Zero marginal benefit. This is the baseline we'll use to see if we can make an improvement. World 2: The amazing world of MIMEzine, the mail reader that sucks out your brain into the computer. A $500 value, but available to you at no charge from your friendly ftp site! Note that there is no monetary exchange in either of these worlds. I want to make it perfectly clear that transaction costs are usually non-monetary, even though they are, in a strict sense, paid. In standard bad old economic analysis, the mail reader is a good (i.e. worth something) that is available for no cost, and so clearly would be used by everybody, because it's in everybody's best interest to do so. As Coase pointed out, not so fast. In order to accurately assess the economic effect of this transaction, you have to look at the whole thing, from start to finish. Here is a not so outlandish sequence. Some of the following costs can be shared between multiple transactions, some can't. 1. Which friendly ftp site has MIMEzine? Make an archie query. Cost: time to make an archie query 2. How do I use archie? Find out by reading the documentation. Cost: time to read documentation and figure out how you'd actually use it. 3. How can I possibly find out what ftp site has the file? Have someone tell you Use archie. Cost: time to ask your computer friends, which you've spent a long time cultivating. [See note below on this topic.] Alternate Cost: $25-$40 for one of those internet books. 4. Download MIMEzine using ftp. Cost: see above for archie, and analogize. 5. Compile MIMEzine for Unix. (Binary distributions need not apply.) Cost: Ever ported? 6. Learn how to use MIMEzine. Cost: time to read manual. time to correct screwups created by inadvertent use of your previous mailreader's keyboard bindings. time spent hunting for instruction on how to set up "proactive filter mocking", which you just have to use. 7. Customize MIMEzine for you own environment. Cost: time to learn what all the little configurations options do. time to choose a place in the directory structure. time to twiddle until you've got it just right. 8. With probability p=3/4, decide that you absolutely can't stand MIMEzine because of some braindead misfeature that makes you less productive or because it's not really compatible with everything else you're using. Cost: multiple all the preceding costs by 4=1/(1-p) to reflect that you keep trying packages until you find one you like. In my own experience, I think a multiplier of 4 is on the low side. 9. The benefits of using MIMEzine! Benefit: Savings of an hour or so a week handling your email. Increased ability to handle content types you're not really interested in. Transaction costs are _all_ of the costs above, since, of course, the package is free, or rather, free(?). To summarize: World 1: The status quo. Often acceptable. World 2: The new technology. Frequently an extreme time sink for what you get out of it, even if it's free software. Is it any wonder that software progresses slowly? A note on friendship networks. The need to have a network of friends that you use to find out about computer stuff is an indicator of serious lack of scalability in the technical and social design of computer systems. Not everyone has time to cultivate a techie network, and most people don't. This indicator is both a design criterion and a test. One should design software so that it can be used without needing to ask question, and one can guage success in this by seeing the number of questions that are actually asked. There is much more to be said about categorization of transaction costs and what can be done to alleviate them. Later. Eric