>From: [EMAIL PROTECTED] >Date: Sat, 4 Mar 2000 02:42:33 -0500 (EST) >To: [EMAIL PROTECTED] >Subject: BOUNCE [EMAIL PROTECTED]: Non-member submission from >[[EMAIL PROTECTED]] > >From: [EMAIL PROTECTED] >Date: Fri, 3 Mar 2000 23:40:32 -0800 (PST) >X-Sender: [EMAIL PROTECTED] >To: [EMAIL PROTECTED], [EMAIL PROTECTED], [EMAIL PROTECTED] >Subject: Re: [pws] New Networking Primitives for Mac >Cc: [EMAIL PROTECTED] > >Je77 and Iannis: > >Than you for the VERY rapid VM testing. With this kind of timing >problem, it's hard to say for sure when it will crop up. Just because >it hasn't yet happened on the G3 tower, doesn't meant it might not >happen there once in 100 times. But then again, it could be that OS 9 >fixed a bug in the underlying networking code. If you do further testing, >please keep me informed! > >I assume you are using the new "acceptFrom:" mechanism that (like >Unix and Windows) allows you to specify a queue length for incoming >connections. This is supposed to prevent back-to-back connections >from being missed. However, you could still miss connections if the >queue length was shorter than the maximum burst of connection requests. >And, of course, if the server doesn't do "accepts" fast enough, you >could also exceed the capacity of the queue. > >Actually, one possibility just occured to me. The Powerbook does >a couple of things involving the disk that are premptive--Squeak >(and all other applications) stop completely while these operations >occur. One case is when the hard disk has spun down and you try >to access the disk. The other case is more peculiar. It happens a >few seconds *after* you've last written something to the disk. >If you start a "BouncingAtomsMorph", you can see when these pauses >happen. The pauses are long enough that a LOT of connections could >arrive during the pause. You can set the power-saving mode so that >the hard disk doesn't spin down, but I know of no way to disable >the pause that happens a few seconds after writing. I'm not even >sure what it is doing, although I suspect that the OS is parking >the disk heads to reduce risk of a hard disk crash if the computer >is jolted suddenly. > >Do your observations fit this theory? > > -- John > > >At 10:08 PM +0100 3/3/00, Ioannis Zannos wrote: > >About the refused request-timing problem on Mac > >with APWS/Comanche: > >This problem occured on a 400 MHz, G3 Powerbook. > >It disappeared when I installed the same Swiki with > >an identical configuration on a 400 MHz G3 tower > >running MacOS9. So I still do not know > >what is the problem. In both cases I was > >accessing the swiki via Netscape from the > >same machine as it was running on. > >I will report when and how the problem > >gets solved on the Powerbook. It is intriguing. > > > >Iannis Zannos > >SIM > >Berlin (D) > > > >"Jochen F. Rick" wrote: > >> > >> John Maloney, from Squeak Central, is changing around the networking > >> primitives for Macintosh. We have tried his new VM so far on seaweed. > >> seaweed is a PowerPC 603 100MHz running MacOS. Where before we >could not get > >> any of the images to load without a reload, the images loaded fine on > >> the first try. We stress tested it a little and found that under high load > >> it would occasionally refuse a connection, but I think that's to >be expected. > >> > >> Anyway, I'm not sure if he wants to make it public yet, so I'll leave > >> that up to him. Thanks John. > >> > >> Peace and Luck! > >> > >> Je77 -------------------------- Mark Guzdial : Georgia Tech : College of Computing : Atlanta, GA 30332-0280 (404) 894-5618 : Fax (404) 894-0673 : [EMAIL PROTECTED] http://www.cc.gatech.edu/gvu/people/Faculty/Mark.Guzdial.html
