In the output of sysctl -oa I see values like: kern.proc.all: Format:S,proc Length:75264 Dump:0x00030000000000002011fac380d4fec3... kern.proc.proc_td: Format:N Length:75264 Dump:0x00030000000000002011fac380d4fec3... kern.file: Format:S,xfile Length:18616 Dump:0x340000003e0c0000e803000006000000... vfs.nfs.diskless_rootaddr: Format:%Ssockaddr_in Length:16 Dump:0x00000000000000000000000000000000... debug.hashstat.rawnchash: Format:S,int Length:262144 Dump:0x00000000000000000000000000000000...
, etc., and some like: kern.ipc.msqids: Format: Length:3520 Dump:0x00000000000000000000000000000000... kern.ipc.sema: Format: Length:600 Dump:0x00000000000000000000000000000000... and even some like: machdep.consdev: Format:T,struct cdev * Length:4 Dump:0x00000000... The first group has widely nonuniform "Format" fields, the second don't have it at all, and the third just looks wrong (is that a NULL pointer being exported from the kernel, just in case the userland doesn't have it? :) ). My proposal is to MIME-ify the Format fields, best presented in examples: "S,proc" -> "x-struct/proc" "S,xfile" -> "x-struct/xfile" Etc. GEOM tree is the easiest, it's just "text/xml", and the already present text fields would be "text/plain". If possible, the types should conform to http://www.iana.org/assignments/media-types/ - but except for text and XML types they usually won't. The next step, obviously, but of dubious benefit, would be to stop exporting binary data from the kernel and do it via XML, but that's another thing, possibly another GSoC proposal :) I can provide the patches for the format names. At this time I'm looking for input: is this idea sane? Are there any utilities that actually parse the type name? The benefit of this conversion is that MIME types are easier on the eyes and will help unfamiliar users understand what's going on.
signature.asc
Description: OpenPGP digital signature
