declan@portia.stanford.edu (Declan McCullagh) (07/15/90)
Howard Chu writes, quoting my earlier message... >I'd definitely take a NeXT over a Mac (I have, in fact }-), but ... >until *this* [System v2.0] is delivered, > >I wouldn't recommend either to anyone else. Running find on my winchester >(ostensibly looking for core files, but there are lots of valid reasons) >brings the entire system to a crawl, if not a halt. The display manager >takes upwards of 20 seconds to respond to window/mouse selections, and you >can watch window redraws one widget at a time. This is pretty sad performance >for a machine that supposedly has mainframe-style I/O channels and a 25MHz >data bus. I've noticed the same slowness of response when doing very disk-intensive work. How about launching applications? Try running TopDraw on an 8 MB NeXT and time how long it takes to load. And now try running a large program on even a Mac SE, and notice the difference. Of course, the reasons for this slowness are "valid", and NeXT consciously chose this path when they adopted their development environment... Just try doing something in the foreground when you're launching an application, and watch your perceived response time take a nose dive. This problem doesn't seem to be nearly as bad on a 16 MB cube, though... (Just another reason to add more memory!) Even so, I wouldn't seize on poor application launching time as a reason to tell people not to buy a cube - after all, once you've got your app launched, you've got your perky interface back. And to go back to your "find" example - there's nothing stopping you from nice'ing the process up a bit... >A simple getty sitting on an open (floating, temporarily disconnected) serial >line eats up huge amounts of CPU time when it *should* be sitting idle. But >for some reason, it prefers to spit out "Next Login" endlessly, cutting in to >my compile performance. I've got a little RS232 tester connector on the line, >it shows that DCD is low, and that getty is set on a dialup device, but it >always succeeds in opening the device. Pretty poor. (And this stupid Mac >mini-din serial connector drives me nuts too. No ring indicate, no RTS/CTS flow control, what a stupid idea.) Agreed, serial ports seem to be the NeXT's weak spot. I STILL don't have my modem working properly on dial-in, through no fault of DTS - they've been very helpful... I have the same infinite "NeXT Login (ttyx)" problem, too. The reason it's cutting into the performance of the cube is because every time those RD/SD lights flash on your EIA-232 tester, a new getty process is created, which sends those characters to the serial port, and then init forks a login - which eats up a considerable bit of processor time if this happens every three seconds or so. And I'm still waiting for SLIP to be released. >>Personally, I don't want to have to trade in MY greyscale monitor for a color >>one - I want BOTH on my desktop. >> >You either have a huge desk, or a paperless office, or maybe both. }-) >My current desk is a fair size, but just barely deep enough for the monitor >and keyboard. (Hey, I guess we need voice or eye-tracking input too, eh?) Let's just say that if I had the opportunity to have a multi-monitor'd NeXT system, I wouldn't let simple lack of desk space stand in my way... >Oh, horrors, then they might have to actually provide a primitive power switch >I think this soft-power switch stuff is nonsense. At least on my good old HP41 >it made sense - you could program the thing to turn itself both off and on. It >doesn't seem to serve much good on the NeXT, and it's damn frustrating when th >machine goes off the deep end. (Then I have to crawl around down dark corners >looking for the plug...) Of course, I have a similar gripe about MacIIs. So it >goes... (Don't know what it is, but these things tend to crash pretty often in >my experience. Am I cursed?) I think I've only encountered a situation where I HAD to pull the power switch twice in my experience with the cube - and these seemed to be Window Server related (although I can't prove it) $-). The command-key reboot sequence works quite well, although it's not the nicest way to do things (remember Ctrl-Open-Apple reset on Apple II's?)... >There's nothing preventing any other company from creating really slick softwa >bundles for their systems; Apollo had a pretty nice package for their 2500/350 >products. And there's nothing to prevent anyone from adding some whiz-bang >floating point unit to their workstation; the DSP doesn't need to be tightly >coupled to the CPU, it wouldn't be a difficult add-on. The other guys may take >a long time realizing there's a demand for DSP-equipped workstations, but it's >no great hardship to supply the rest. Most optical drives are SCSI compatible, >after all... Of course, there's nothing preventing any other company from creating "really slick software" for their own machines, but NeXT has a tremendous advantage. First, the development environment is demonstrably superior to anything I've seen under X Windows; second, they've had the opportunity to start from scratch with their interface and don't have to support an existing base of GUI apps; third, they've been at it for quite a while, and the others are just starting to think about having a "real" user interface that doesn't force you to drop down to the command line every few seconds... Also, even though the 56001 is a relatively inexpensive chip (<$50), it's still much cheaper for a company to include it with their motherboard than to add it in as a separate processor board later. I've been looking at DSP boards for the Macintosh, and I haven't seen one for less than $2-3,000... Does anyone have the pinouts for the Optical Drive connector inside the cube? Is it SCSI? I'd love to try to hook it up to my Mac... > -- Howard Chu @ University of Michigan -Declan ------------------------------------------------------------------------------ Olympic Technologies / Registered NeXT Developers \ declan@portia.stanford.edu ------------------------------------------------------------------------------