I actually have seen that site, I think it's pretty funny. I don't take it very seriously, I got a good laugh the first time I saw it, although I assumed it was someone poking fun at Gentoo users in general. On further consideration, I think your characterization actually may represent a larger component of the community than I'd like to admit, perhaps? I think I'll try to remember that me and my associates using Gentoo may not be a representative sample. I didn't really mean it as an attack, sorry if it came off as one.
I take it you were just trying to make a couple of points, that there are a lot of uninformed Linux users, to which I completely agree. And also, your earlier (funny) point about the USE flags and such, I find myself in agreement with that as well.
It counter-productive to over-generalize and stereotype, and actually that "9 out of 10" statistic sounds suspiciously made up. Go talk to some folks on freenode's #gentoo about it, I believe you'll find that most of them just wanted a level of control that you simply can't get from something like Fedora or Ubuntu.
Nobody I know really encourages rediculous optimizations with Gentoo. All the users I know are professionals, or are trying to become more experienced. I would ask that you cite a source to support your "ricer" comment.
I can add a couple more to that list. After spending a few hundred hours playing base assault and domination from CoD United Offensive, imagine our suprise when they were missing from CoD2.
Still a bit disappointed about it, as you can probably tell.
I agree with everything you said here. I actually thought VLIW/EPIC had a lot of promise once. My only problem with Itanium is that it was used as a justification to kill off so many other promising designs, a couple of which I thought still had life left in them at the time.
While this is perhaps a bit off-topic, I found this funny:
I imagine that many users find it particularly annoying that if they want to, say, create a new spreadsheet, they must launch OO.o, which puts them in Writer, then go to File->New Spreadsheet in order to get to Calc.
Hey, why not run "oocalc" instead of "ooffice"? That opens up the spreadsheet program instead, at least in 1.1.4 that I'm using. I'll leave it as an exercise to the reader to discover oodraw, ooimpress, oowriter, etc.
After playing quake for years, I tried to play UT2k3 and I just felt like I was stuck in molasses.
Which is why I always played it at 1.35 speed, in mostly low gravity... I think almost everyone agrees with you that the apparent speed of 2k3 was too low.
Sequential file access detection could probably be done, if not in the general case, then in a large number of cases.
However, I think a better solution is to add code to those userland applications to use the already available mmap()/madvise() (MADV_SEQUENTIAL or MADV_DONTNEED, as mentioned before) or even open() w/ DIRECT_IO (bleh), and assume the kernel will do the right thing. The applications are in a better position to make this determination than the kernel.
I've actually had to fight the perception that swap causes bad behavior on every unix variant I've administrated. Once in awhile, people do notice some bad behaviors, but most of the time, the system will just adapt to the new workload, given time.
I'm not sure why this is even a thread. If you don't like swap on your desktop machine, turn it off (or down), in one of the several methods available. I'm keeping mine.
And if many different companies ran IM servers that were all part of the same system, how long before SPIM became rampant? Right now, AOL/MSN/Yahoo can shut down accounts that violate their terms of service.... how would that be handled if other companies connected into one IM system? My guess is that it would become as bad as usenet or email is now... except IM demands your immediate attention. Imagine getting 100 extra IM s an hour for Viagra and low, low mortgage rates.
And with clients like gaim and trillian, why does this matter to anyone except the competitors?
Also remember: the SGI machine we're talking about is a NUMA architecture, which means that the software will need to be written for it. Unlike classic SMP (like the Sun Fire), it has a kernel image for each CPU
There seems to be a great deal of confusion about NUMA around here. There is no reason to rewrite any software, this is not a cluster using MPI or something similar. The number of times I've seen this falsehood repeated is distressing.
It's possible that the kernel text resides on each node within this NUMA box (IRIX does this on the Origin 2000, at least) for performance, but we're still talking about a single system image.
Routers (R-bricks) (and meta routers, if they're still used in larger configurations) make the memory on all nodes appear local, the only difference is that latency off-node is higher, depending on the number of hops.
Slashdot Mirror
Mine must be overclocked, I get 392.40. :)
There's a sucker born every minute, I guess. Example from the "article":
The horizontally enlongated layout gives the site to much Air, causing instability.
Shut it, dirty hippy. I think I'm getting a bit stabby.
I actually have seen that site, I think it's pretty funny. I don't take it very seriously, I got a good laugh the first time I saw it, although I assumed it was someone poking fun at Gentoo users in general. On further consideration, I think your characterization actually may represent a larger component of the community than I'd like to admit, perhaps? I think I'll try to remember that me and my associates using Gentoo may not be a representative sample. I didn't really mean it as an attack, sorry if it came off as one.
I take it you were just trying to make a couple of points, that there are a lot of uninformed Linux users, to which I completely agree. And also, your earlier (funny) point about the USE flags and such, I find myself in agreement with that as well.
I completely agree with you, but I would also add that their bugzilla instance has helped me immensely, as has IRC (irc.freenode.net #gentoo).
It counter-productive to over-generalize and stereotype, and actually that "9 out of 10" statistic sounds suspiciously made up. Go talk to some folks on freenode's #gentoo about it, I believe you'll find that most of them just wanted a level of control that you simply can't get from something like Fedora or Ubuntu.
Nobody I know really encourages rediculous optimizations with Gentoo. All the users I know are professionals, or are trying to become more experienced. I would ask that you cite a source to support your "ricer" comment.
I can add a couple more to that list. After spending a few hundred hours playing base assault and domination from CoD United Offensive, imagine our suprise when they were missing from CoD2.
Still a bit disappointed about it, as you can probably tell.
I agree with everything you said here. I actually thought VLIW/EPIC had a lot of promise once. My only problem with Itanium is that it was used as a justification to kill off so many other promising designs, a couple of which I thought still had life left in them at the time.
While this is perhaps a bit off-topic, I found this funny:
http://projects.csail.mit.edu/gsb/archives/old/gsMowing? Why didn't you say so:
t m
http://www.friendlyrobotics.com/NewSite/robomow.h
A friend actually has one. Part of his loyal robot army.
You saw that? It was just the one time.
I imagine that many users find it particularly annoying that if they want to, say, create a new spreadsheet, they must launch OO.o, which puts them in Writer, then go to File->New Spreadsheet in order to get to Calc.
Hey, why not run "oocalc" instead of "ooffice"? That opens up the spreadsheet program instead, at least in 1.1.4 that I'm using. I'll leave it as an exercise to the reader to discover oodraw, ooimpress, oowriter, etc.
After playing quake for years, I tried to play UT2k3 and I just felt like I was stuck in molasses.
Which is why I always played it at 1.35 speed, in mostly low gravity... I think almost everyone agrees with you that the apparent speed of 2k3 was too low.
OpenOffice.org will have a fix ready within days, but how quickly will Linux users patch?
However long it takes emerge to finish. Duh.
Not to be alarmist, but Missouri has had it share of earthquakes though. What is probably the largest in the continental US:
h tml#february_7
http://neic.usgs.gov/neis/eq_depot/usa/1811-1812.
Proves nothing.
m &q=l&c=RHAT
http://finance.yahoo.com/q/bc?t=5y&s=SUNW&l=on&z=
Sequential file access detection could probably be done, if not in the general case, then in a large number of cases.
However, I think a better solution is to add code to those userland applications to use the already available mmap()/madvise() (MADV_SEQUENTIAL or MADV_DONTNEED, as mentioned before) or even open() w/ DIRECT_IO (bleh), and assume the kernel will do the right thing. The applications are in a better position to make this determination than the kernel.
I've actually had to fight the perception that swap causes bad behavior on every unix variant I've administrated. Once in awhile, people do notice some bad behaviors, but most of the time, the system will just adapt to the new workload, given time.
I'm not sure why this is even a thread. If you don't like swap on your desktop machine, turn it off (or down), in one of the several methods available. I'm keeping mine.
And if many different companies ran IM servers that were all part of the same system, how long before SPIM became rampant? Right now, AOL/MSN/Yahoo can shut down accounts that violate their terms of service.... how would that be handled if other companies connected into one IM system? My guess is that it would become as bad as usenet or email is now... except IM demands your immediate attention. Imagine getting 100 extra IM s an hour for Viagra and low, low mortgage rates.
And with clients like gaim and trillian, why does this matter to anyone except the competitors?
Also remember: the SGI machine we're talking about is a NUMA architecture, which means that the software will need to be written for it. Unlike classic SMP (like the Sun Fire), it has a kernel image for each CPU
There seems to be a great deal of confusion about NUMA around here. There is no reason to rewrite any software, this is not a cluster using MPI or something similar. The number of times I've seen this falsehood repeated is distressing.
It's possible that the kernel text resides on each node within this NUMA box (IRIX does this on the Origin 2000, at least) for performance, but we're still talking about a single system image.
Routers (R-bricks) (and meta routers, if they're still used in larger configurations) make the memory on all nodes appear local, the only difference is that latency off-node is higher, depending on the number of hops.
Absolutely. There is a newly expired draft of an RFC for the OpenIM standard here