I want to start out by disclaiming that yes, I don't know much about Irix yet. I'll never know as much about Irix as I do about Linux. I also admit that I know little about *really big* installations of UNIX (and its competitors); except that they're so far removed from your average desktop user's installation (even UNIX desktop), that it's a whole 'nother world. On Wed, Nov 21, 2001 at 03:30:51PM -0600, Jeremy Shawley wrote: > On Wednesday 21 November 2001 10:40, Carl Wilhelm Soderstrom wrote: > I agree with this statement to a point. Security is alwasy a concern for any > company. Having open accounts after install I think is pretty insane myslef. > In the whole password area IRIX is really lacking. Just the pure fact that > IRIX is still using crypt passwords and not MD5 is behond me. does TrustedIRIX use anything other than crypt()? or is TrustedIRIX's philosophy more along the lines of 'account for what everyone does so we can blamestorm when the brown stuff flies'? (That's not a criticism of TrustedIRIX; but rather of government logic, which TrustedIRIX was designed for). > The whole ssh thing is a bit convoluted. Because we deal so much with > government, ssh has to be an afterthought. IRIX may never include ssh by > default, but ya never know. what does gov't have to do with it? (other than 'govt = !"common sense"'). > > - user command line tools are horribly neglected. AFAIK, no shell that it > > ships with, has command completion (certainly not by default). > /bin/tcsh > Command line completion. but for those of us who would rather use bourne shell syntax... ;> > BTW, you can get bash from freeware.sgi.com :-) yep. got it. just hate to go through that extra step. for that matter, the extra hassle of grabbing bash, gnu-ls, and vim is enough to put me off using the *BSDs more; so if you want to level the criticism of 'lazy bastard' at me, it's justified. :) (it's one of the good qualities for an admin). > There is a logical reason for home directories being in /usr/people, that is > the fact you are better off having an nfs mounted filesystem in a sub > directorie than in a root location. So if /usr/people becomes unavaliable, > you won't have trouble access / If I were to have the filesystem mounted as > /people , and the nfs server for /people went down, I will have timeouts and > slow access the / partition. ok. valid point. I always thought it was there for historical reasons (left over from the days when /usr was for _users_); but that's probably a decent reason to keep that. the way we get around that here at Real-Time is have the automounter mount home dirs individually; so /home/carl will be mounted, no matter what box I'm logged into (behind our firewall at least), but no one else's would be. I admit that this may not scale well to systems with 1000s of people logged in simultaneously; but those systems are few & far between, and I don't see that it's sensible to penalize the majority of people, for the sake of the exotic cases. for the exotic cases; build an exotic config. (tho I suppose one could justify that IRIX deals with the exotic cases more often than just about any thing else). > I agree that some tools are in very strange > places. Why ping is kept in /usr/etc/ I will never know. There is also a good > reason for having freeware put in /usr/freeware. SGI has some tools that > can't be replaced by freeware version. Perl is the best example. If you were > to replace SGI's perl with the one from freeware, then things like nsd would > not work. BTW, there is a script in /usr/freeware/bin called "fixpath" that > will add /usr/freeware/bin to your local path. so why not put them in /usr/local/ ? > > - a simple installation of it, takes up nearly 800MB. that's GUI, > > compiler, and a lot of user tools. > And how is that any larger than a default linux install? My last base redhat > workstation install was over a gig. yeah, RedHat has really pigged out over the last few years... I've seen Debian workstations (Nate's laptop) which were happily under 500MB. (Including compiler, X, and the fancy wallpaper. ;> ) I've got a NetBSD box with gcc and basic utilities (including bash and vim), no X, under 70MB. in any case, you're right; OSes are all horribly bloated these days, and linux distros are rarely an exception. That doesn't mean I can't gripe about it. :) > > - (more of a hardware issue) that fancy graphics hardware does *so much* > > in hardware, that flexibility is compromised. a single graphics 'pipe' > > (which may be multiple monitors) is limited to 8 million pixels in its > > framebuffer; and no more than 3840 pixels in any one direction (horizontal > > or vertical). my desktop at home is wider than that (4320 pixels); and IBM > > now has a single monitor that requires a graphics card capable of 9.2 > > million pixels. > > If you are refering to the demo you had at the sgi office, the system you > were looking at was over 4 years old. I thought the graphics dude said that the new systems still had the same limitation? I asked him about it. > I think a better undrstanding of how > graphics hardware works would be a good class to take. oh, absolutely. I think a lot of the reason that people don't appreciate what SGI hardware can do; is that most software they see, couldn't take advantage of it anyway. Games would have much more realistic scene rendering; with much more object vertices, more realistic curves, and much better texturing. ever notice that the first-person shooter games mostly take place indoors? and when you're outside (Tribes 2); there's few/no trees, and much of the action takes place on flat plains? PC 3D hardware can't keep up with rendering even simple trees in a halfway-realistic forest. SGI stuff can; and scale it to run on *multiple monitors* (which I *really* wish more games could do). > How you can even > compare a single nvidia card to an onyx 3000 is behond me. I'm not. :) for the fancy 3D stuff, SGI equipment still blows everything out of the water, and I fully expect it to keep doing so. I'm just amazed that there are limitations on how many pixels wide you can draw a framebuffer. Can the X implementation 'join' multiple pipes together, so you don't need to write your graphics app to be multipipe aware? > Here is some > better information for you. Lets see, you can hang up to 6 monitors off of a > single pipe, have tools like ircombine to put them together and have > functional multipal monitors. Time to flip on Mr. Coffee and do some > research. > > http://www.sgi.com/realitycenter/ very sexy hardware. the place where IRIX wins here, is the X implementation being able to 3D accelerate across multiple heads. XFree86 can't do that yet; tho maybe next year we'll see something (that's a wild-ass guess). the commercial X servers for linux can do that, tho. where IRIX and Linux diverge heavily, is at this point. As I understand it, Linus resisted including direct-graphics-rendering code in the Linux kernel; because it violated one of the basic tenets of his OS design, that only the kernel should be allowed to bang directly on the hardware. IRIX has had direct rendering for a *LONG* time, because that's what seems to be needed for fast graphics work. (OS and hardware designers may disagree; and I'm certainly interested; but that's another discussion). IRIX has a *far* better direct-rendering implementation at the moment (AFAIK). Linux's is still very much in its infancy. Linux is also contending with the hurdle of all the different hardware implementations out there; whereas SGI only has a limited set of hardware to deal with (and it's better-designed hardware at that). so even if you ported SGI's X implementation to Linux, and taught Linux how to bang on that hardware; it would still lose to IRIX, at least for the next couple of years. :) > let's not even get into > who's kern is better. When linux has things like systune, sar by default, depends what you mean. there isn't a unified tool yet for reading and tweaking kernel parameters; but once you learn the /proc filesystem, hdparm, and a few other tools, most of the pieces are there. as for sar: chrome at steel:~$ apt-cache search atsar atsar - system activity reporter but you're right; there are probably a lot of system admin tools that IRIX has, and most linux distros don't include by default (and linux admins haven't heard of, so they don't know to ask for workalikes). > and a simple kernel recompile (on IRIX all you need to do is autoconfig -vf > and you have your new kernel). what do you mean 'kernel recompile'? you don't have the source code, do you? the advantage that IRIX has, is that it runs on a much more limited set of hardware; and it's better-understood hardware. lots of linux drivers and architecture-understanding is done by guess and by golly, since the manufacturers are (understandably, at times) unwilling to tell us exactly how it works. > I'm not trying to be flame bate here, but I think for a real comparison, a > person needs good experience on both linux and IRIX. no, that's a perfectly valid point. that's why I'm asking questions; because I honestly do want to know more about IRIX. > I don't think that linux will ever be > able to touch IRIX in its stability and scailability, I actually beg to differ on that point. it may be a while; but 'ever' is a *very* long time. I think in 4-5 years, there will be few things that IRIX can do, that Linux won't be able to. there's nothing magical about it; it's just evolution. Linux is the fastest-evolving OS out there; and the rate is accelerating. > (please checkout some of the apps that came with your box > like moviemaker. videoscope, soundeditor, SGI Meeting, and also the > documentation (ivv at the command line). ) is wonderfull. yeah, I've looked at some of that stuff before, and it's really great. :) > Sorry for being > such a ass about much of this, I just get tired of the uneducated trying to > educate the uneducated. If you don't know something, than just admit it. I think I tried to do that, maybe not well enough. :) If we don't try to throw some discussion out there tho; and have the bad points refuted, we'll never learn. (just look at MS users [even NT admins]... how much do you see most of them talking about other OSes?) :) > I > admit there are many linux things that I'm not very up on and some IRIX > things that I am rusty at. (Notice how I did not comment on the xlv > question). well, that's why I didn't say anything about it myself. for all I know, it's twice as cool as Linux LVM; but I don't have the faintest clue about it. > But to compare something that you have tinkered with to something > that you have deicated your life to is just not a fair comparison. no, it's not a fair comparison. most things in life aren't. you have to make comparisons based on heuristics, and when possible gain more information to refine your heuristics with. :) hey, I didn't blow away the IRIX install on that Indigo2 and install Linux or NetBSD did I? :) > Anything but 4Dwm as a default window manager. the main gripe that I have with commercial unices, is the Not Invented Here syndrome. everyone has their own licensed version of the Bourne Shell, the Korn Shell, the C Shell, vi, X, etc... Linux does the cool thing and says "hey, so-and-so has a nifty tool that does everything we need it to do; let's not reinvent the wheel"; and the people making that decision, are (more or less) the end users. so the system gets refined to be more and more like what most end users want; and not what some marketdroid decides based on 16 million pages of research. <flamebait> We can even see this in Linux distros... compare RedHat to Debian. Redhat is still a commercial company; and there are managers who decide what would be good to put in the distro, that would benefit Redhat Inc. Debian is entirely user-driven; and stuff goes into the distro that *users* want. This is why Debian has apt, and Redhat does not. Redhat wants to sell you update CDs and subscriptions to their update service; Debian wants to distribute the update process as much as possible, because that makes it cheaper and more convenient for people. </flamebait> Carl Soderstrom. -- Network Engineer Real-Time Enterprises (952) 943-8700