ted@sri-spam.UUCP (01/27/87)
Tannenbaum (sp?) was at Usenix last week, and held a BOF (Birds of a Feather) session on Minix. I thought I'd pass on some of the things I remember. Anything wrong is doubtless my fault; I wasn't taking notes. In no particular order: The system supports the notion of split I and D space, as V7 did on pdp 11s. However, because of a limition in either the compiler or the assembler (which were munged from existing versions), only small model code can be produced (I and D share one 64K area. This corresponds I think to the .com model in MSDOS). The utilities supplied in compiled form were not compiled in the same environment and some of them DO have split I and D. The file system is not vanilla V7, in particular there is a bitmap for free blocks rather than a free list, and only one time is maintained in inodes rather than the 3 kept by V7 (and u*x in general). I think this has serious implications if you want to scale the system up to the point where the "dump" command would be nice to have, since it keys off the inode modified field to allow true exact restorations of file system state. Most applications (make ls etc) only care about the last update time for a file, and will have the info they need. Tannenbaum admits that due to its implementation as a message passing system, Minix will never set any speed records. The person who wrote the compiler driving tables for the 8088 hated the chip and did a mediocre job. Recompiling the system and utilities will result in somewhat larger and slower code. There is no serial driver to get at the serial ports. The hooks are there in the tty code to support rs232 terminals and communications, but it has not yet been done. There are people working on this. The system can not yet boot from a hard disk. The Atari ST port is underway, but apparently supporting the 68000 with its nice linear address space is harder than supporting a segmented architecture when there is no memory mgt. No support for graphics hardware. Problems with some AT hard drives. Ted Nolan -- ------------------------------------------------------------------------------ | | | Ted Nolan UUCP (Probable): | | 575-B Winding Creek Rd. ..ucbvax!ucbcad!nike!sri-spam!ted | | Fayetteville NC, 28305 ..seismo!caip!nike!sri-spam!ted | | ..akgua!usceast!ted (long forward) | | Internet: | | ted@braggvax.arpa | | ted@sri-spam.arpa | | | | "Feather the rast!" | ------------------------------------------------------------------------------