Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!samsung!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!emory!hubcap!retrac From: retrac@titan.rice.edu (John Carter) Newsgroups: comp.parallel Subject: Re: Paper (and) systems Keywords: Distributed Shared Memory for-real Message-ID: <8108@hubcap.clemson.edu> Date: 21 Feb 90 21:40:07 GMT Sender: fpst@hubcap.clemson.edu Lines: 46 Approved: parallel@hubcap.clemson.edu In article <8075@hubcap.clemson.edu> af@spice.cs.cmu.edu (Alessandro Forin) writes: >When compiling bibliographies, it would be most helpful if one asked such >questions as > 1- Has the system been implemented for real ? > 2- Is it available to external sites ? > 3- Does anybody actually use it for serious work ? > >I am sure this would reduce the size of such bibliographies by various >orders of magnitude. It would also improve their utility by the same >factor or more, esp on lists such as this one. It really depends on what you want the bibliography for. If what you're interested in is the current research going on in the field, then you want to know about *all* the work that's been done or proposed so that you can decide for yourself what are the good ideas and what aren't. You wouldn't really care if the people with the "neat idea" have spent the of man hours necessary to clean up the user-interface, remove the last few bugs, write up the nice user documentation, and otherwise make the idea "commercial quality". That's the difference between the R and the D in R&D (research and development). Of course, it *is* useful to know which systems/algorithms/whatever *are* "commercial quality", but not so that you can discard the rest as so much trash. For example, Apple commercialized the ideas that were developed at Xerox PARC for mouse-driven, window-oriented user interfaces, but that certainly doesn't make Apple's "research" the only stuff worth having in your bibliography! Quite the opposite, in fact. >As far as I know, the bib on DSM would contain, at this time, only one entry: >Mach. And Linda, if we use the most general definition. > >I'd be happy to be convinced there are more. As far as I know, Mach is the only system that claims to meet your three requirements stated above. However, I certainly don't think that the only research in the area of distributed shared memory that is worth reading about is that associated with Mach. I hope that you don't think that, either. Leaving Kai Li's Ivy system out of a bibliography on DSM just because it isn't in wide circulation is silly. Research != Development. -- John Carter (retrac@rice.edu -or- {any-Internet-node}!rice!retrac) Rice University, Dept of Computer Science, Houston, TX Go OWLS! Go ROCKETS! Go DOLPHINS! Go REDS!