Path: utzoo!mnetor!uunet!husc6!bbn!mit-eddie!bloom-beacon!gatech!mcnc!decvax!ucbvax!ucsd!sdcsvax!ucsdhub!hp-sdd!hplabs!sdcrdcf!trwrb!desint!geoff From: geoff@desint.UUCP (Geoff Kuenning) Newsgroups: comp.graphics Subject: Re: Yet more elementary dross.. a suggestion Message-ID: <1712@desint.UUCP> Date: 9 Apr 88 06:08:41 GMT References: <1705@desint.UUCP> <10253@steinmetz.steinmetz.ge.com> Reply-To: geoff@desint.UUCP (Geoff Kuenning) Distribution: na Organization: Interrupt Technology Corp., Manhattan Beach, CA Lines: 83 In article <10253@steinmetz.steinmetz.ge.com> snyder@boreas.UUCP (Derek Snyder) writes: > Ever hear of inter-library loans? I'm not from California, but many libraries > here on the east coast will forward material, or in the case of journal > articles, copies of the articles. If you can find a reference (most university > libraries should be able to help you with a literature search also), > and your library doesn't have the item, chances are they can get it for > you. As it happens, all of the libraries in the UC system, plus some others like USC, are tied together this way; there's even a computer database (I think it's called Euclid) for searching through all libraries at once. Using this, it is possible to locate all articles with certain keywords in the title, or by a certain author, plus subject-categorized searching. Assuming I live in a built-up area like LA (I do), the system then works like this: (1) Drive 45 minutes to UCLA (the nearest one for me), and use Euclid to look up the keywords I can think of. Pay $100-$200 for a library card (or, cheaper, enroll in any extension course and pay $8.75) on the spot, and order the 5-50 articles that the search discovers. (2) Wait about a week while these things show up. Of course, they won't show up all at once (some will be out), and the library won't hold them forever until you come to get them. So you will have to make 2-3 more trips to see what you ordered. (3) Of course, keywords and subject categorizations always miss a lot of stuff, as well as producing a lot of chaff. (The last time I researched something, my first search missed an *entire journal* devoted exclusively to the subject I was interested in.) So you will have to go through the references and develop a supplementary list of things to order. Often, this will be larger than the first. (4) Repeat steps 2-3 several times, resulting in a total of 5-10 trips to your not-so-nearby library, to get a comprehensive list of references on the subject. Each trip takes 1.5 hours of your time, not counting time spent in the library. (5) Decipher all the articles you collected. Many will be mere expansions on previous articles, saying "we assume the reader is familiar with the Foo algorithm and notation", so that you have to track down and read a 1965 article before you can figure out the thing you have isn't relevant. (6) Having selected a promising algorithm or algorithms, repeat steps 2-3 at least once more, this time getting all copies of the next year or two of the relevant journal after the article appeared, so you can search for corrigenda and letters entitled "A Note on the Foo Algorithm." Typically, these items aren't indexed, yet they are often of critical importance. (7) Now the fun begins. Implementing the algorithm can take weeks to months, depending on complexity and on how well the paper you selected describes it. Ever try to produce decent C code for a 1975 Fortran-specified algorithm? And that's the easy case; it's even more fun working with a paper translated from a foriegn language that never gave code in the first place. I know whereof I speak; I have researched more than one subject exhaustively, starting essentially from scratch. The subject I covered most thoroughly literally took me many months and occupied several hundred hours of my time. Given the difficulties of this (which are badly compounded if you aren't close to a good university library), is it surprising that people ask the net? A simple net broadcast has a good chance of reaching a person who makes a specialty of the field you are investigating; that person often can point you directly at the 2 or 3 really important works on the subject. An example is compression: there are probably 50 papers in the field, but all you really need to read is Lempel-Ziv. Furthermore, there is a good chance that you will discover somebody who has working code and is willing to send it to you. > But making some effort to find > the answer to your question yourself is probably more beneficial to you > and less annoying to the rest of the network. Depends on whether you want to know about the subject, or just working code. If I plan a PhD thesis on the subject, I'd agree. But if the problem is incidental to my main goals, I'm interested in a solution, and the quickest way is often the net. -- Geoff Kuenning geoff@ITcorp.com {uunet,trwrb}!desint!geoff