|
eschaton posted:another thing they took from NeXT preach it
|
#
?
May 2, 2015 22:18
|
|
|
|
| # ? Apr 25, 2024 14:02 |
|
|
eschaton is my favorite poster
|
|
#
?
May 2, 2015 22:19
|
|
|
floptical am I right?
|
|
#
?
May 2, 2015 22:25
|
|
|
Wild EEPROM posted:in your job as a clown posting about clown computers maybe. thats not my job
|
|
#
?
May 2, 2015 22:39
|
|
|
eschaton posted:another thing they took from NeXT it was a super nice UI, but it was a dogshit unix on hilariously overpriced, underpreforming hardware they really never had a chance
|
|
#
?
May 2, 2015 22:42
|
|
|
Notorious b.s.d. posted:a dogshit unix on hilariously overpriced, underpreforming hardware in a time when you had a wide array of choices for just that experience
|
|
#
?
May 2, 2015 23:24
|
|
|
Notorious b.s.d. posted:it was a super nice UI, but it was a dogshit unix on hilariously overpriced, underpreforming hardware you've never actually said why you think NEXTSTEP was “a dogshit Unix” — it was pretty much just a 4.2BSD like SunOS or Ultrix, but in a colocated kernel with Mach 2.5 for its VM and task/threading/IPC. (and using its own binary format instead of a.out or COFF, which meant it had shared libraries from the beginning, instead of as some afterthought.)
|
|
#
?
May 2, 2015 23:27
|
|
|
eschaton posted:you've never actually said why you think NEXTSTEP was “a dogshit Unix” — it was pretty much just a 4.2BSD like SunOS or Ultrix, but in a colocated kernel with Mach 2.5 for its VM and task/threading/IPC. (and using its own binary format instead of a.out or COFF, which meant it had shared libraries from the beginning, instead of as some afterthought.) you just explained for yourself why it was a dogshit unix: lovely kernel, non-standard toolchain, lousy compiler. being colocated on mach 2.5 means it's literally hundreds of times slower than the competition. poo poo got better in the 90s with INKS, but not by much you also named literally the only unix worse than nextstep: ultrix. another crappy compiler, another vendor who doesn't care, and no shared libraries. (just running xterm and emacs at the same time could kill an ultrix machine) being nominally "based on bsd" doesn't mean everything is equal. SunOS 4 was miles and miles ahead of ultrix or nextstep Notorious b.s.d. fucked around with this message at 23:38 on May 2, 2015 |
|
#
?
May 2, 2015 23:35
|
|
|
what was the standard toolchain? suncc? xdb?
|
|
#
?
May 2, 2015 23:36
|
|
|
Subjunctive posted:what was the standard toolchain? suncc? xdb? in a nextstep context, 4.3bsd's crap
|
|
#
?
May 2, 2015 23:41
|
|
|
just in general nextstep was not a very compelling package it cost as much as a macintosh or a unix workstation, while being slower than either and not really compatible with either set of software packages it was a mac-like play years too late with no major ISVs on board cool UI experiment, awesome hardware design, really fuckin dumb as a product to sell to real human beings who exist
|
|
#
?
May 2, 2015 23:44
|
|
|
Notorious b.s.d. posted:you just explained for yourself why it was a dogshit unix: lovely kernel, non-standard toolchain, lousy compiler. except the kernel wasn't lovely, there was no standard toolchain—you even admit further on that basic features like shared libraries weren't common yet—and the compiler was actually pretty good since the one CPU for which gcc actually did decent codegen was 68K. quote:being colocated on mach 2.5 means it's literally hundreds of times slower than the competition. poo poo got better in the 90s with INKS, but not by much you really don't know what you're talking about here, do you? you're just spouting things you vaguely remember hearing about research versions of Mach. NeXT did not use a Mach as a single- or multi-server, do syscall-via-IPC, etc. NeXT went for straight collocation, on the advice of people who had worked on Mach like Avie. this resulted in performance comparable to Sun on similar hardware. (NeXT even used Sun3 hardware to bootstrap.) go ahead and dis the performance of an 8MB system running off and swapping to a magneto-optical the speed of a 1x CD-ROM all you want, that was a dumb idea. but the kernel and userland were both decent and got perfectly reasonable performance on similar hardware to what Sun and Apollo and HP and SGI were shipping at the time. quote:being nominally "based on bsd" doesn't mean everything is equal. SunOS 4 was miles and miles ahead of ultrix or nextstep in my experience NEXTSTEP was not significantly different from a userland or porting perspective from SunOS 4 on Sun3. it just had a way, way better UI and its own shared library implementation (which every Unix did then, if it even had shared libs).
|
|
#
?
May 3, 2015 00:34
|
|
|
Aftershocks of the epic nix flamewars of the 90s.
|
|
#
?
May 3, 2015 00:39
|
|
|
theyre still epic 
|
|
#
?
May 3, 2015 01:34
|
|
|
eschaton posted:you really don't know what you're talking about here, do you? you're just spouting things you vaguely remember hearing about research versions of Mach. i flatly do not believe this i never spent time benchmarking nextstep, i've never even seen nextstep benchamrks, but osf/1, an os people actually cared enough about to benchmark, was brutally slow, for the same reasons as nextstep. mach 2.5 + 4bsd is just not an ideal kernel eschaton posted:go ahead and dis the performance of an 8MB system running off and swapping to a magneto-optical the speed of a 1x CD-ROM all you want, that was a dumb idea. but the kernel and userland were both decent and got perfectly reasonable performance on similar hardware to what Sun and Apollo and HP and SGI were shipping at the time. it was a dumb idea, and they were charging $10,000 for it cool ui, lovely product
|
|
#
?
May 3, 2015 02:19
|
|
|
Notorious b.s.d. posted:i flatly do not believe this too bad, doesn't matter what you believe, it's how things were. quote:i never spent time benchmarking nextstep, i've never even seen nextstep benchamrks, but osf/1, an os people actually cared enough about to benchmark, was brutally slow, for the same reasons as nextstep. if OSF/1 was slow due to Mach, it would be because they were trying to use the single-server or multi-server pattern, with IPC instead of syscalls. that would be very different than, not the same as, NEXTSTEP. quote:it was a dumb idea, and they were charging $10,000 for it that's why they abandoned it the following year when the pizza box NeXTstations were announced. the 040 cube only had an optical drive as an option. they were cheaper than a comparable Mac (a Quadra 700). and performed similarly, too, because Mach as used by NeXT just did not impose the kind of overhead you seem to believe Mach always does.
|
|
#
?
May 3, 2015 04:09
|
|
|
pram posted:eschaton is my favorite poster
|
|
#
?
May 3, 2015 05:48
|
|
|
how does it all end?
|
|
#
?
May 3, 2015 07:38
|
|
|
duhhheschaton posted:best Unix then, best Unix now in the form of OS X, the world's most advanced operating system
|
|
#
?
May 3, 2015 07:52
|
|
|
Gnome 3 question: is there any way I can make the screen turn off for power saving after x minutes without locking the screen? I cannae find the button to do that anymore.
|
|
#
?
May 3, 2015 08:18
|
|
|
Rahu posted:Gnome 3 question: is there any way I can make the screen turn off for power saving after x minutes without locking the screen? I cannae find the button to do that anymore. Settings > Privacy > Screen lock
|
|
#
?
May 3, 2015 08:29
|
|
|
rip jochen liedtke l4 was super duper fast microkernal
|
|
#
?
May 3, 2015 12:49
|
|
|
page 219 of an operation system thread containing operation system discussion
|
|
#
?
May 3, 2015 14:31
|
|
|
Progressive JPEG posted:page 219 of an operation system thread containing operation system discussion
|
|
#
?
May 3, 2015 14:51
|
|
|
219 page of linux on the yospos
|
|
#
?
May 3, 2015 14:58
|
|
|
i finally found a nextstep benchmark, but it's on hppa instead of black hardware. it's a set of 3 graphics benchmarks from the 90s. the numbers are milliseconds. lower is better. all of these were done with varying software configurations on an HP 712/60, the "cheap" HP-UX workstation that was supposed to kill the PC code:
|
|
#
?
May 3, 2015 16:36
|
|
|
i think a big problem with using those graphics numbers as a software benchmark is that it's as much a test of the window system as the unix system underneath i think it is a real possibility that the nextstep window system is just that much slower than hp x11. at least, that would be my guess for the 80%+ performance differences in the center column.
|
|
#
?
May 3, 2015 16:59
|
|
|
the yospos number...
|
|
#
?
May 3, 2015 18:59
|
|
|
Notorious b.s.d. posted:i think a big problem with using those graphics numbers as a software benchmark is that it's as much a test of the window system as the unix system underneath yeah, Display PostScript was definitely slower than X11, but still plenty fast for end-user UI purposes. my mono slab certainly feels decent running OPENSTEP 4.2, though OmniWeb is admittedly slow. (it wasn't so bad over dialup in 1997 though…) I've been trying to find BYTE Unix Benchmark numbers for the original cube (25MHz 68030) to compare to the Sun 3/80. The closest I've found are NeXTstation (25MHz 68040) numbers, which have it performing somewhat better than a SPARCstation IPC on the whole. maybe BYTE benchmarked A/UX 3 on a Quadra 700 or one of the final HP/Apollo 68040 machines, for a more like vs like comparison.
|
|
#
?
May 3, 2015 19:09
|
|
|
eschaton posted:yeah, Display PostScript was definitely slower than X11, but still plenty fast for end-user UI purposes. my mono slab certainly feels decent running OPENSTEP 4.2, though OmniWeb is admittedly slow. (it wasn't so bad over dialup in 1997 though…) pretty much anything will outperform an ipc. it had a 25 MHz SPARC v7 -- that means no hardware multiply/divide. even the 68k-based 3/80 was faster. it seems like the easiest like-for-like comparison would be ns vs real unix on 486/sparc/hppa, but then you run into how horrible gcc is, as in those graphics benchmarks Notorious b.s.d. fucked around with this message at 20:10 on May 3, 2015 |
|
#
?
May 3, 2015 20:08
|
|
|
Notorious b.s.d. posted:pretty much anything will outperform an ipc. it had a 25 MHz SPARC v7 -- that means no hardware multiply/divide. even the 68k-based 3/80 was faster. the best like-for-like would be a benchmark of another 25MHz 68040 Unix system like an HP 9000/425. nobody's scanned the ancient BYTE with a review of that model though. though I bet someone running Linux or NetBSD on an 040 has posted benchmark scores...
|
|
#
?
May 3, 2015 20:22
|
|
|
eschaton posted:the best like-for-like would be a benchmark of another 25MHz 68040 Unix system like an HP 9000/425. nobody's scanned the ancient BYTE with a review of that model though. though I bet someone running Linux or NetBSD on an 040 has posted benchmark scores... the baseline system for the byte unix benchmark is a sparcstation from the early 90s. people only ran it on old systems for shits and giggles, which is probably why it's hard to find results. nobody was running this benchmark when the hp9000 400 series was a current product i did find some circa linux 2.0 results from a 486 and they were pretty loltastic linux was slower than nextstep 3.x. an impressive accomplishment
|
|
#
?
May 3, 2015 20:29
|
|
|
eschaton posted:another thing they took from NeXT i'm just quoting this to point out this ui is way, way, way better than osx is today nextstep was wildly non-standard and encumbered by postscript and often run with monochrome graphics but man was the user experience good
|
|
#
?
May 3, 2015 20:31
|
|
|
Smythe posted:the yospos number...
|
|
#
?
May 3, 2015 20:31
|
|
|
Notorious b.s.d. posted:i'm just quoting this to point out this ui is way, way, way better than osx is today that's difficult for me to imagine because i've never used it and it just looks like a slightly better version of windows 3.1
|
|
#
?
May 3, 2015 20:33
|
|
|
it is somehow appropriate that we are discussing the first actually-good unix desktop on page 219 those monochrome nextstep shots from 1988 are especially impressive when you consider that the competition was this, released in 1987:  sun microsystems thought this loving shitshow was an appropriate thing to unleash on the world, in 1987. this was their second attempt at a gui. the first attempt was even worse.
|
|
#
?
May 3, 2015 20:35
|
|
|
hm for some reason i thought steve was forced out of apple a lot later, like 91/92.
|
|
#
?
May 3, 2015 21:02
|
|
|
Notorious b.s.d. posted:it is somehow appropriate that we are discussing the first actually-good unix desktop on page 219 Subject: yo!
|
|
#
?
May 3, 2015 21:06
|
|
|
linux literally looks like this, to this very day
|
|
#
?
May 3, 2015 21:09
|
|
|
|
| # ? Apr 25, 2024 14:02 |
|
|
unix desktops, then and now, are plagued by the fact that most of the users mostly want to arrange a bunch of terminal windows. if the graphical ui is half-baked no one really cares, as long as the core task of organizing xterms works out there are no normal, non-developer users. never have been. this makes gnome 3 especially laughable because it fails at the one thing that actually has to work. i would rather use suntools (stools?) from 1985.
|
|
#
?
May 3, 2015 21:29
|
|