eugene@wilbur.nas.nasa.gov (Eugene N. Miya) (08/16/90)
Dave Sill proposes a comp.benchmark.
This might be a good idea, but in my opinion, only if the group is moderated.
It would serve no other purpose than existing groups if not otherwise.
There is a performance mailing list and it's only use to date has been
to call for papers or announce TRs. It should be more than a platform
for people to announce what they have.
Dave should be or find a willing moderator, but as a LAST resort, I'd do it.
Comments about iron handed moderation aside.... 8^)
--e. nobuo miya, NASA Ames Research Center, eugene@orville.nas.nasa.gov
{uunet,mailrus,other gateways}!ames!eugeneaglew@dual.crhc.uiuc.edu (Andy Glew) (08/17/90)
>Dave Sill proposes a comp.benchmark. >This might be a good idea, but in my opinion, only if the group is moderated. Sorry, Eugene, I strongly disagree with you here. Maybe if it was split up into two newsgroups, comp.benchmarks.theory => moderated comp.benchmarks.practice => unmoderated. I want, and hope, for a "practical" benchmarking group, one where I can ask questions like "I have version 1.27 of the gravsim parallel benchmark running on an Encore, and it is core dumping after N thousand iterations. Does anyone have any idea what might be going wrong" (to which hopefully someone could say "Yeah, I had a similar problem on a Sequent. You set the -DBAZZ define wrong"). Something like this is important. I've spent the last week working around things like that. Moderation adds too much delay for this sort of interaction. Conversely, a moderator would be useful for the more staid side of benchmarking: reporting new results, theoretical questions, announcements of availability, and so on. But I doubt that there would be enough traffic to make it worthwhile... ---- Flash! I just realized that what the USEnet needs is not moderated newsgroups, but recommendation tags applied to various news messages. I mean, a moderator's function, apart from slowing things down, is to remove inappropriate messages (important), but also to add some sort of filtering effect - eg. if Bob Levine thinks it's worth reading, I probably should read it. Now, if all newsgroups were unmoderated, and messages were distributed immediately... But if the present day moderators, after seeing and approving a message, sent around a tag that subsequently got attached to the stored message, saying "Moderator MMHGH approces of this message, and thinks its worth reading" Then we could have the best of both worlds. Problem: forging of approval tags. Solution: same mechanisms of trust currently used. Problem: approval tags could occupy a lot of space. Potentially every reader of the USEnet might want to send out approval tags. Solution: Filter. A system need only accept approval tags from the "official" USEnet moderator(s) for a group. Problem: potential size increase of groups that are now moderated. Solution: apply early expirations to not-yet-approved messages. Only accept messages that already have an approval tag (which is equivalent to moderation). And so on. -- Andy Glew, a-glew@uiuc.edu [get ph nameserver from uxc.cso.uiuc.edu:net/qi]
eugene@wilbur.nas.nasa.gov (Eugene N. Miya) (08/18/90)
In article <AGLEW.90Aug16185619@dual.crhc.uiuc.edu> aglew@dual.crhc.uiuc.edu (Andy Glew) writes: >I want, and hope, for a "practical" benchmarking group, ... > Something like this is important. I've spent the last week >working around things like that. Moderation adds too much delay for >this sort of interaction. > >Conversely, a moderator would be useful for the more staid side of >benchmarking: reporting new results, theoretical questions, >announcements of availability, and so on. I don't know, I think you can have that. I think Steve Stevenson does a pretty good job moderating comp.parallel, and Peter Neumann on comp.risks. I do not think the delay is that critical. It is not just a function of time but also quality. >Flash! I just realized that what the USEnet needs is not moderated >newsgroups, but recommendation tags applied to various news messages. Perhaps. I'd like to see removing lots of the "me, too" postings, cutting down the attribution (editing), especially when attributing signatures. People do not really summarize what's transpired. Practical is benchmarking. The theory is MVA, queueing theory, simulation and emulation, in performance analysis. I want empirical stuff, too. --e. nobuo miya, NASA Ames Research Center, eugene@orville.nas.nasa.gov {uunet,mailrus,other gateways}!ames!eugene
seanf@sco.COM (Sean Fagan) (08/19/90)
In article <AGLEW.90Aug16185619@dual.crhc.uiuc.edu> aglew@dual.crhc.uiuc.edu (Andy Glew) writes: >comp.benchmarks.theory => moderated I *like* that idea. I would assume it would be a place to discuss benchmarks in general, e.g., what makes a good one, how to improve specific ones, etc? >comp.benchmarks.practice => unmoderated. Hmm. I have this picture of vendors posting results of new compilers / systems to this, and find that vision both appealing and nauseating... 8-) In that case, I would expect it to be called "c.b.results"... -- Sean Eric Fagan | "let's face it, finding yourself dead is one seanf@sco.COM | of life's more difficult moments." uunet!sco!seanf | -- Mark Leeper, reviewing _Ghost_ (408) 458-1422 | Any opinions expressed are my own, not my employers'.