From mboxrd@z Thu Jan 1 00:00:00 1970 To: 9fans@cse.psu.edu Date: Tue, 11 Jul 2000 15:09:16 +0000 From: "John S. Dyson" Message-ID: <8kfa9b02tqg@enews3.newsguy.com> Content-Type: text/plain; charset=us-ascii References: <8kf0s012ck0@enews3.newsguy.com> Subject: Re: [9fans] Any significant gotchas? Topicbox-Message-UUID: d7c7880c-eac8-11e9-9e20-41e7f4b1d025 In article <396B1F62.24083.11FD481@localhost>, ngr@9fs.org (Nigel Roles) writes: > > >> As some of you'all *might* know, I like playing with >> OSes... Any major gotchas in a planned booting >> of plan9 on my SMP box? :-). (Dell Precision 620, >> dual 866 Xeon, with 1GB.) >> > > > THe usual gotchas are > > 1) video card > 2) SCSI card > > read the hardware compatibility list. > Of course... How is stability? Is there opportunity for significant improvements in the kernel? If I choose to run 1000 (or 10000 or 20000) processes on a single machine (not uncommon on some machines nowadays), will the kernel work well, degrading smoothly? If I might want to spend alot of time working on the code, I'd like to know where you think that the improvement opportunities are. When I initially looked at the code a few years ago, I was impressed with both the simplicity, but also saw some scalability problems. It was really great to see code that hasn't already been obscured with complexity. From an immediate gratification standpoint, this might be alot of fun for me. It is well understood that relatively less used OSes will often need more driver support than they already have. So, I kind-of already knew of that issue. Note also, there are often Network Card driver issues -- some manus don't always fully document their cards, and change PCI ids, thereby confounding probe code. This problem is common, and thankfully, there are alot of source-available OS codes available to work from. Also, one sometimes finds that X86 SMP implementations, with differing APIC layouts, confound kernel startup. I am not initially depending upon running SMP, but it would be interesting to hear anecdotal reports. If the silly thing doesn't boot up, on my day-to-day OS development machine, I'll buy another machine with no hesitation. It might be a good idea anyway to buy a Plan9-only machine for any initial work. (For example, FreeBSD required that I recompile the kernel so as to support the large number of INTRs available on the Dell 620. Things dont always work out of the box, even on slightly more common OSen.) I have a copy of the entire comp.os.plan9 for the last few months, and am encouraged with the traffic level. John