Re: proposed Eistring interface

I see I did get closer this time. :-) Ben> i still disagree because it can corrupt the innards unless Ben> you expect automatic binary conversion? Then someone who Ben> feeds in JIS-encoded data this way will not get Japanese, but Ben> raw data, and perhaps shouldve used eicpy_ext(). It can't corrupt the innards because it does do binary conversion. Of course Japanese should be forbidden there, as should _all_ encoded text. Only when the programmer is willing to take reponsibility for correctly handling the encoding at the raw octet level should the ei*_c APIs be used. Eg, in mswindows_get_file() you handled a similar issue of correct formatting by checking for a trailing `\'. But when someone needs to insert ISO 2022 escape sequences by hand, or such like, the ei*_c APIs are convenient, and they also signal that the programmer is doing something potentially dangerous. Not to XEmacs; but to the external system or to the user's data. That's true of mistakes with ASCII, too. The following is an aside and not directly related to Eistrings anymore. Ben> by its nature, the default int. rep. must be able to Ben> represent all chars. that would rule out utf16 if we have Ben> more than 1,000,000 and some chars. but it doesn't rule out Ben> ucs4, or some utf16 extension that could encode gigs o' Ben> chars, etc. OK, I asked because I wondered if you'd thought about it in detail. Apparently not. Here's how I see it. We do have that many character position codes if we admit UCS-4 or UTF-8. But there will not be such a UTF-16 extension; Unicode is committed to staying within the UTF-16 character set for the forseeable future. There is already a UTF-32 specification for a 32-bit wide Unicode representation; this is also committed to the 17*65536 UTF-16 code space, and will not be expanded beyond that for the forseeable future, despite the potential for 2^31 (they are also committed to ISO-10646 compatibility) characters. However, there are uses for the huge UCS-4 range. Tomo and his buddies are already playing with the so-called "konjaku-mojikyo" pseudo-charsets, which have over 70,000 code points already assigned. These are quite popular with Japanese Windows users, too. Anyway, they will eat up most of the UTF-16 private space if used in the obvious way. Nor do we want to encourage "Japanese exceptionalists" to borrow not yet-standardized-parts of the UTF-16 space. Also, although _we_ should not support "language-tagged character" encodings (pace, Olivier) by default, we should permit third party libraries and extension modules to do so. This could be easily done using UCS-4 private space. Before Hrvoje jumps in, let me say that given that the Unicode standard will not expand out of the UTF-16 range, I don't think konjaku-mojikyo or other linguistic research purposes are good enough reasons for not using UCS-2 as a default internal representation. I will think about how to handle that potential conflict in a reasonable way, to forstall the abuses that I would expect from Japanese programmers. We should, of course, support --with-default-internal-representation={ucs4,utf8} for people who want those huge spaces. The issue here is how to have XEmacs gracefully decline to handle them when the default internal representation is UCS-2, not such a big deal. -- University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN Institute of Policy and Planning Sciences Tel/fax: +81 (298) 53-5091 _________________ _________________ _________________ _________________ What are those straight lines for? "XEmacs rules."