home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!olivea!hal.com!darkstar.UCSC.EDU!osr
- From: jesup@cbmvax.cbm.commodore.com (Randell Jesup)
- Newsgroups: comp.os.research
- Subject: Re: Lightweight threads and "real" users
- Message-ID: <1833h8INNe57@darkstar.UCSC.EDU>
- Date: 2 Sep 92 19:09:28 GMT
- References: <13cn3eINN8cn@darkstar.UCSC.EDU> <13sb8pINNn28@darkstar.UCSC.EDU> <17n06qINN793@darkstar.UCSC.EDU> <17tq4mINNfdl@darkstar.UCSC.EDU>
- Organization: Commodore, West Chester, PA
- Lines: 22
- Approved: comp-os-research@ftp.cse.ucsc.edu
- NNTP-Posting-Host: ftp.cse.ucsc.edu
- Originator: osr@ftp
-
- fouts@bozeman.hpl.hp.com (Martin Fouts) writes:
- >In article <17n06qINN793@darkstar.UCSC.EDU>, Jack_Palevich@taligent.com (Jack Palevich) writes:
- >|> 3) Between address spaces use synchronous send-and-wait-for-reply
- >|> messaging, rather than asynch messaging.
-
- >If you don't, there will be a class of applications you will be able
- >to write more efficiently than otherwise. The above rules are good
- >rules and should be followed unless you can discover a compelling
- >reason not to follow them. The trick here is that the reason must be
- >compelling.
-
- I agree, Rule 3 is iffy at best. Programs that have realtime response
- problems would have to hide the synchronicity themselves by spawning even
- more threads. Another good example is an interface manager that sends messages
- indicating UI actions that have taken place (such as "he clicked here"). You
- _don't_ want the UI manager to block because the program was calculating
- an FFT. This is what gave people so much trouble porting things to the MacOS:
- they had to sprinkle GetNextEvent() calls in code that might run for a long
- while, to keep the UI running.
-
- I would think that might be rewritten "use synchronous messaging
- whenever there isn't a good reason to use async".
-