On 8/8/05, Yanek Korff <[EMAIL PROTECTED]> wrote: > > I think there's an important point in the heart of what Piotr (Rybicki) > was saying. Perhaps we don't need to go so far as to test whether BSD > admins know about a variety of system calls, but it may be worthwhile to > test at a slightly higher level. That is, questions that answer: > > * What ARE system calls? > * What system calls might relate to I/O operations? > * What happens when a process forks? (What system call is involved? duh) > * What system calls might relate to file descriptors? > * What's the difference between and fork and an exec? > * What system calls are involved in network operations? > * How are semaphores used in interprocess communication? > * What is a context switch? > > Basically, you'd be trying to see if the administrator could wrap their > hands around ktrace and use it effectively. There's also a better > chance that the administrator will understand what he or she is looking > at under top, iostat, sockstat, netstat, etc. > > There's a bunch of entry-level and 2nd year university C stuff here, and > it may not be very BSD specific either. Perhaps questions about ktrace > itself may be more effective. > > -Yanek.
Hi, all! been lurking for a while... I think Yanek gets at the heart of the issue. Certainly a sysadmin needs to be able to interpret ktrace/kdump output and create effective systrace policies. And that requires a certain understanding of what the calls do. It's also important to be able to customize, patch, and debug the kernel and drivers. And it's important to be able to understand diffs when you're performing upgrades, or evaluating a rooted system. I also think that there is a perhaps valid assumption that admins on F/OSS systems should be prepared to do a little hacking when releases break things. Furthermore, in my experience C and Make are inextricably linked in many people's minds, and there's no doubt that sysadmins need to know Make inside and out. Most sysadmins also need to sort through C macros at some point to make includes point to something someone installed in a nonstandard location. No, a sysadmin doesn't need to be a programmer, but an effective sysadmin does need to understand some programming basics. My litmus test would be this: if you can understand the man pages for OpenBSD system calls, you know as much C as you need, but you need at least that much. Again, that's a question of basic literacy, not programming skill; it's the difference between taking two years of a foreign language in college, and actually being fluent. There is also the question of what makes this a specific BSD certification, and not just a general sysadmin certification like cSAGE. Sure, there are BSD variants of commands, and some BSDs have specific functionality like pf and systrace, but for the most part system administration is much of a muchness on all non-windows platforms. The difference needs to be something more substantial than knowing that that hostname doesn't have an -i flag on BSD. One way would be to offer a module on system internals, or device drivers. Just my $.02, -- jay -------------------------------------------------- This email and attachment(s): [ ] blogable; [ x ] ask first; [ ] private and confidential daggerquill [at] gmail [dot] com http://www.tuaw.com http://www.dpguru.com http://www.engatiki.org _______________________________________________ BSDCert mailing list [email protected] http://lists.nycbug.org/mailman/listinfo/bsdcert
