When the LLG created rafsi, they were assigned partly on the basis of how often the corresponding natural-language equivalents of the gismu were used in compounds. That is, gismu corresponding to concepts that are often used in compounds in natural languages, would get many rafsi, and gismu corresponding to concepts that almost always appear alone in natural languages, would get no rafsi.
However, it soon became apparent that Lojban usage did not follow natural language usage, so the rafsi assignments were not as useful as they were meant to be. So in 1993, the LLG generated a statistic of how often and in which positions the rafsi appeared in lujvo, and on that basis proceeded to assign new rafsi to the gismu, in order to make as many lujvo as possible as much shorter and less crunchy as possible.
- The above isn't quite correct. The initial set of rafsi was generated the same way as the 1993 generation. However, since Lojban did not have any lujvo yet, we used the TLI Loglan set of lujvo including a vast set of proposed lujvo. TLI lujvo making was not nearly as semantically rule-based as Lojban lujvo-making has turned out, so many of those proposals would not be acceptable in their Lojban equivalents. In any event, by 1993 we had some actual usage of the various gismu/rafsi in lujvo and it seemed appropriate to do a revision prior to baselining the rafsi list (the gismu list had been baselined in 1992). I did a complete reanalysis using the same rules I had used originally. The community overrode this analysis, saying that some chunk of rafsi were so well known that they were "sacred" and unchangeable merely because of a possible statistical advantage. The rafsi actually changed ended up being about half of the proposed set of changes. Bob LeChevalier
- Unfortunately, the reallocation did not result impartially for the set of cultural gismu (a set which I consider inappropriate anyway, though). bemro got three rafsi, while friko got two, despite the fact that bemro is already a subdivision of a continent and thus relatively more specific than friko, which encompasses a whole continent and thus likely has more potentials than bemro as a lujvo element for sub-concepts of the generic "African". I also find it incovenient that cosmetics-/fashion-related words like cinta and skaci have no short rafsi from which to create various derivates useful even in everyday conversations especially among women. tijlan
A comparison between the old and the new rafsi list can be seen here: http://www.lojban.org/files/history/bothraf.txt
This has the side effect that Lojban written prior to 1993 looks funny to people who are reading it now. For example, xisli'icutci (rollerblade, xislu linji cutci) has survived in the lujvo list for many years, even though by today's rafsi assignments it means xriso linji cutci! The Lojban spoken between 1987 and 1993 is sometimes poetically referred to mijyjbo, "Middle Lojban", in reference to a poem by Michael Helsem (see LojbanicMythology).