About The Commodore Knowledge Base

The Commodore Knowledge Base is an assimilation of various articles on Commodore repair, programming and technical issues, ranging from beginner level to advanced, covering a diverse series of topics from CP/M to hardware modifications. CKB covers the gamut of Commodore 8-bit systems, not Amiga and very little on the PC series. Emulator content is intentionally limited, although cross-platform development is a considered topic. It is administered by Cameron Kaiser and sponsored by Computer Workshops, Inc. Hosting provided by Floodgap Systems.

CKB articles originate from many sources, largely the Usenet group comp.sys.cbm, patrolled by a news robot maintained by Cameron Kaiser. Kurt Brandon has also provided a significant number of articles gleaned from BBSes and the now defunct QLink online service. Other notable contributors include: Marko Mäkelä, Stephen Judd, John Iannetta, Doug Cotton, Nicolas Welte, Jim Butterfield, Jim Brain, Ray Carlsen.

Searching With Commodore Knowledge Base

Previously, CKB was powered by the KnoBs system, a simple brute-force search system. KnoBs was a flexible and powerful searching system, but as articles increased in size and number, the KnoBs architecture was scrapped due to speed and CGI overhead reasons. KnoBs' direct successor is Textil, a high-speed index and search engine. Textil adds relevancy rankings, enhanced keyword processing (even limited natural language queries), and rapid response time.

For each match that Textil returns to your query, it assigns a score. This score is based on how common your keyword is in the set of documents, and what proportion of each document your keyword makes up. CKB takes the average and standard deviation for each, and initially returns only the matches that have a score greater than the mean (when you ask for less similar matches, CKB gives you the whole set). Scores are only relevant for a certain set of keywords; you cannot compare the document scores between two dissimilar searches. The higher the score, the more relevant the document is.

Because of the way Textil attempts to determine how relevant a keyword is when it indexes the document, in some cases words that may only appear once or twice in larger documents may be dropped. This is a difference between KnoBs and Textil that increases the engine speed at the expense of a few dropped words. If you believe this is seriously impacting the engine's accuracy, please mail the maintainer (this is tunable).

Downloads

Certain articles (marked with downloadable in the article listing) have embedded programs that the Knowledge Base can tokenise and send you.

To download embedded listings, click the link that precedes them. The very first time an embedded listing is downloaded, it must be unpacked and tokenised, and this can take as long as a minute on longer programs. CKB runs on an overworked computer, so please be patient.

Once you've downloaded the program, it goes into a cache so that other users won't have to wait. Periodically, this cache may be cleared as programs are rewritten or the database is being reindexed.

Embedded programs are still very experimental. Please tell the maintainer about difficulties (corrupted listings, etc.) It does appear that most of the bugs, finally, have been fixed.

Search Commodore Knowledge Base | Mail the maintainer