You may be right..but Shouldn't it be 3 people? I mean, how does it help that 2 people have differing numbers? You still need to spend time to figure out which is right. Now if you had three, the two that had the similar input should be the correct one. The occurence of all three being different is significantly less than 1 person being wrong.
There were no explanation in the book. But I think that in those dark days, where "data entry" was a fact, and a profession, that a lot of hard data was gathered about errors in data entries. I don't have the book anymore, but believe that a single person would enter data with around 88% accuracy. The number seems low, but most entry data was probably long (for the typist) meaningless sequences of digits, like Fortran code, scientific measurements etc.
I guess that two people would enter data, a third person would compare data, and recheck the original source if any difference was found. This would allegedly produce a 97-98% accuracy. One could probably improve this somewhat by adding a third data entry person, but it would increase costs, even if one got rid of the "rechecker", and automatically selected the "two of three must right" data, since the rechecker would only have to check the discrepencies, not the whole data set. So all in all, it is probably a cost-benefit analasys, that decides what one should do. (in the case of this story, I guess a IBM service contract would be the best:-)
I worked with a historian, doing a statistical analasys of approx. 25.000 pages of historical, criminal records. We only took notes of a few things in each record (type of crime, date, age, sentence etc), but it was gruelling work to do.
I had hacked up a data entry form, but didn't make too much out of the data validation. But after examining the data harvest for the first few days, I decided that real anal data validation was needed, since all kinds of data was wrongly typed, or entered in the wrong fields.
It was not because we were sloppy and we actually took an interrest in the job, but hours and hours of turning pages, gleaning information, type and tab just does make you prone to make mistakes.
I read in some old statistical handbook, that in order to gain 97-98% correct data entries, _two_ people should enter the same data.
I guess that this large retyping of 25.000 tax forms, where each form only seems to be typed in and checked by one person, will be full of errors. I wonder how many thousand record that will needed to be redone.
As I understand it, there are enviromental nasty stuff in several of the components that constitute a typical pc. So pc's aren't so easely, and inexpensive disposed of. So there is a problem. A short digression: Almost all enviromental pollution are caused because, those that pollute finds it cheaper to dump their pollution where others must pay the bill, usually taxpayers or property owners, or just live with the consequenses.
Another factor regarding pollution, and its cleaning up, is, that it is cheap to pollute, but expensive to clean up. And since most pollution is done with a economic motive, those that pollute, rarely pays the bill.
So I believe, that economic incentives, like enviromental fees, is a good way to reduce pollution (which may benefit us all), is fair (since it should payed by those causing the pollution, not by the public at large), and most importantly; may cause a shift from manufacturing and consuming enviromental harmfull stuff, to, at least, lesser harmfull stuff.
Back to the topic: So if such a fee should be imposed on pc's, it should be done in such a way, that it "hurt" components, that are the most envorimental damageing (eg. cadmium etc), but not those who aren't. Lets suppose CRT's are enviromental nasties, so lets "fee" them. But if some new CRT technology is not, remember not to leverage a fee on that. Or else there will be no economic incentive to change behaviour. It is important to keep that goal in sight.
On most motherboards, there sits a tiny battery. I really don't know whether it is enviromental damaging or not, but lets assume it is, but a slightly more expensive battery technology is not. One can shurely predict, that all motherboard manufactures will use the polluting but cheap solution. But if the polluting solution becomes more expensive by fees, all the manufactures will turn on a dime, and start using the more enviromental friendly, and cheaper solution. Eg. At least older Compaq pc's had ATA-IDE cables, that were way shorter than the industri standard. Alledgely this was because, that when somebody really produces something in bulk, it pays of to save even a few yuans per produced unit.
An email containing a username and password is sent to the registrant a few hours after submission.
As so many others have mentionend, a "few hours" is a very long time. Perhaps not only because people have a short attention span, but because people, the potential costumers, are comparison shopping, and the delay meant that they went elsewere.
I don't know what software your company is making, but lets assume it is something for the desktop user, a piece of software that may have advanteges over the competition, but nevertheless may be easely substituted with something from a competing company. Eg. a small photo editing app.
Somewhere, a proud owner of a new digital camera, wants that kind of software; he goes searching on the web, looks at screenshoots, featurelists and prices, and decides that 3 products looks promising: two of the products are instantly downloaded and tried out, but your app, requires not only a long registration formular, but induces a surpricing two hour waiting period, in which your potential costumer, not only have tried your competitions apps, but may have actually bought them.
Think about going down "Main Street", shopping for a pair of shoes; In one shop, when asking for trying out a pair of shoes, the expedient hands you a two page formular, asking among other things, your phone number, age and job status. After spending 20 minutes filling it out, you are then told to come back in a couple of hours. You then go elsewhere.
Lets assume, that your software is somewhat more expensive, and not an "impulse" buy. Perhaps an unique app, that will help people design better, and faster "foo". Surely, professionals may be more patient. But no, that work dead afternoon, where you potential costumer is searching the web for tools that may make him more productive, may be followed up by 5 hectic days. So if you don't engage your potential costumer when he has time, you can loose an oppertunity. Same thing with the trial period; if your software cost serious money, it probably requieres several hours to test. Most professionals have way too little time at their disposal, they may only have some short timeslots availeably during a week, for testing something new. 30 days may pass quickly, so bump the trial period to 60 days (like eg. IBM does).
In short, make your product as easely availably as possible.
I sure wish they'd figure out how to issue binary diffs instead of complete rpm packages. How much bandwidth was wasted having millions of people download a dozen full packages for the 10 lines of screwed up code in zlib? (No, I don't want to compile it from source. I just want binary packages signed with the disto's gpg key.)
Actually Red Hat does have a binary diff program, called rhmask. Red Hat also used it around RH 5.X/6.0), by issuing "mask" rpms for larger packages like Netscape Communicator.
Even though such diffs seems smart, they trade ease of use, simplicity and less work, for a smaller download size.
eg.: to use a binary diff, you also need access to the original package, which you likely don't have lying around on your server. Then you need to manually generate a new package, and first then, are you able to apply the security patch. More steps to go through, more info to read, more work to do. While the penny-wise modem user may accept that, the overworked server admins don't like it.
So I believe that the rhmask scheme was killed by popular demand.
What about the Red Hat Network? I subscribe 'cause it makes my job as admin SOOOO much easier - but the RHN largely consists of servers with BIG, FAT PIPES.
As I understand it, it won't be easy to sneak a trojan into RHN, just by cracking RH's RHN servers, since all the RPMS are gpg/pgp signed (+md5sum). So an altered rpm-file, will fail the Up2date agents gpg check against RH's public key, and AFAIK therefore not be automatically installed.
So installing Red Hat rpm's, even from "untrusted" mirrors, should be safe, provided that RH can keep their private key secret, and you actually verify the gpg signing (easy to do).
Seems like this also might be an excellent point from which to launch a big DDOS attack, no? How closely does RH watch their servers?
I do think that RH actually monitors their servers (and bandwith), but perhaps much more important, I am quite sure, that they will respond quickly, if people reported a DDoS attack from their IP-range.
IMHO a big part of the DDoS/SPAM/skript-kiddie problem is, that so many boxes, are running totally unattended and / or administrated by less than entirely competent people, who may not even read mail adressed to eg. postmaster@example.com, or without the skills to comprehend the described problem.
A 2 hour DoS is nuisance, a 10 day DoS a disaster.
> I probably won't be prioritizing government traffic on any of my routers.
Definitely his/her perogative. But it's such a common sentiment (not trusting the government) I wonder...at what point *would* you voluntarily help out the government to a good end?
When I read the above quotation, I thought the poster meant, that none of his routers are Internet backbone routers, and therefore the proposals are very unlikely to affect his equipment, since prioritized goverment traffic won't pass through them anyway.
Anyway, most sysadmins would probably be very veary about implementing and open up, any new protocols/features on their routers, unless they really had to. Eg. ECN (congestion control) is probably a good idea, but since so many routers/firewalls/sites still doesn't understand the ECN bit, you can lock yourself out, from perhaps 8-10% (*) of the net, if you enable it. (* quote from some site, doesn't know whether they area good estimate or not)
IMHO, the your box is underspecced (ram, ide harddisks) for the job you are doing. Of course, you can try read: /usr/src/linux[name]/Documentation/sysctl/v m.txt for some tunable/proc parameters (eg./proc/sys/vm/ overcommit_memory should be 0 (zero))
Since you are using ide disks, 'man hdparm' is your friend. Check your kernel config for dma support of your mobo chipset.
Daniel Robbins (from gentoo linux) has written an interesting article "Maximum swappage" http://www-106.ibm.com/developerworks/library/swap tip2.html
Linux allow you to parallelize swap, just like a RAID 0 stripe
Eg.: spread your swapfile on two disks, with equal priority. That way, you should in theory, double RW access speed for the swap. Also, some gains could be gained, if the swap partitions
were moved from disks, that the OS and apps writes to. But read the article.
A lot of people say "Oh, so DU ammo is dangerous, 'snicker'".
When I was a kid here in Denmark, it was a fairly common event, that some 40 year old WWII "horned" mine was seen drifting into a harbour. I remember that a couple of kids that died, because that rusty old tincan they kicked, in fact was a german stick granade. (those "potatomasher" granades are higly unstable).
Even today it is very common, that fishermen gets a stack of corroded gas granades, usually mustard gas, in their nets, since incredible amounts of WWII gas ammmo, was dumped into the baltic sea after the war.
Some years ago, I visited a woodclearing where german small arms and AA ammo was tried destroyed. It wasn't a well done job; the entire clearing was littered with shells. The holes where the detonations had taken place, was still, after 55 years, without a trace of a single leave of vegetation. Probably caused by the phosphor from the tracer rounds.
The danish coastline was part of the Atlantik Wall, and therefore heavely mined (more than 1.4 million mines). Roughly 1 mineclearer died, for evey ten miles of coastlines, and there are still areas not cleared to this day.
The rest of Europe and the former USSR is littered with WWI and WWII ammo. The "war most be fought with all means" proponents, really lives in the "here and now", and forget the decades, and centuries that comes afterward, and the huge amount of civilians who has to live on or near the former battlegrounds.
The Soviet Union did not embrace information technology. It was a society that was essentially living in the 1930's. Heavy industry was the priority of the USSR, not semiconductors.
If you looked on the desks of Soviet desk jockeys in the late 80's, you'd find most offices to be non-computerized (like many western offices). The ones with computers had green screens, IBM or Apple clones. Engineers had Intergraph or Apolla stuff.
The USSR was indeed behind behind the west regarding advanced semiconductor technologi, but your anectdotical evidence can be misleading, since the USSR soviet economy was sharply devided into a civilian part (who got almost nothing) and a military who had first priority. So even though the standard USSR office was pen-and-paper, the military complex would have access much more advanced technology. IMHO, soviet military equipment since WWII to until the eighties, was often on par, if not better, than US equipment (especially missilies, tanks, infantery weapons, airplanes, though perhaps not avionics). OTOH, civilian USSR equipment was always decades behind, what could be found in the west.
The truth isn't bigoted or ignorant. I believe that a famous USSR newspaper was called "Pravda", meaning "The Truth";-).
RedHat hates KDE, that's why they do as few packages as possible. Come on! KDE has been a part of RH distroes since RH 5.2 (?)
Bero (maintainer of the RH KDE rpms) from RH is working very hard on KDE.
He even has a home page with automatically build rpms from daily cvs snapshots.
Their testing style is different, though, and probably more agressive:
KDE 3.0 is found in Red Hats Raw Hide dir. But instead of releasing a whole KDE 3.0 beta 1, then a beta 2, etc, they release batches of incremental improved rpms, often from daily KDE 3 cvs snapshots.
So generaly speaking, RH KDE 3 rpms are much more recent, than those builds found from other vendors on ftp.kde.org.
The upside from this strategy is, that eg. a showstopper bug reported to bugzilla, can be eradicated quickly. The downside is, that a new batch of eg, KDE-networking rpms may break everything, and that everything goes so fast . (eg . I reported a bug known from the changelog, because of that).
One of the reasons, I believe, for RH to sport such an aggressive testing strategy is, that the next RH release will be build entirely with the GCC 3.x
compiler. KDE 3.0 (alpha and first beta at least) couldn't be build with gcc 3.x, and a lot of fixes were probably necessary.
If KDE 3.0 is stable when RH 8.0* hits the market, you can be sure they will include it.
* I think they are going to bump up the version, since an entire distro build around gcc 3.x, is a major step. (compat gcc for all the 2.7x,/2.9x source code).
And yes! cups (www.cups.org) look like it is going in too. A very, very cool printing system. IBM's JFS is going in too, and of course, KDE 3.0
How come there's no mention of 'Terminate', wich was by far the best terminal emulation software for DOS? I still keep all my porn on c:\ter400\download\pictures
Terminate is still around, now in a 5.0 version (www.terminate.com).
I don't claim to have tried a lot of terminal emulation programs, but Terminate was really outstanding.
Heat
Unless the room is broom-closet sized, or you got a lot of equipment (more than 5 or 6 Athlon/p4 servers), the equipment can probably survive without active room cooling. Internal cooling of the cabinets may have to be beefed up, especially multi harddisk systems (cheap to do though).
Perhaps some creativity may help too. Perhaps some of the systems doesn't need to run 24/7.
Some BIOS's have an internal timer and calender, so you can shutdown the systems when likely not in use.
WoL (Wake on LAN) to remote boot, suspend or shutdown systems can be nice too (almost all nics and mobos support WoL nowadays).
Hook it up with some X10 gadgets and a sensor, so that the system(s) boot, if you go near your bedroom console at night, or you alarm clock goes of in the morning, or if you start your coffe machine after 2 o'clock in the night, or...
Other power management features may be present in the OS, so you can suspend the entire system, or just the harddisks, by a cron/at jobs.
Not only will you save some money, but the room will run cooler too.
And unless you run your own DNS, mailserver, etc, then a shut down firewall/router at night, is probably the safest firewall you can get;-)
Dust
This is my nemesis at the moment, our server room is in a basement, with an untreated cement floor.
I suspect our DAT and some other stuff, died because of the cement dust (ok, so DATs always break down after a short while, but..). Anyway, fans and PSU's seems like dustmagnets, which again leads to worse internal component cooling, so a clean room, without carpets is my recommendation.
Noise
All your equipment will make an infernal noise, and a generally bad indoor clima in the room. Of course, people have very individual sensitivity to this, but personally I prefer to hack outside the serverroom.
I final note, if you run a Linux box, then I can only recommend netsaint, from www.netsaint.org.
It is a very flexible, very reliable monitoring system. Since it checks services with plugins, it is easely extensible to include eg. room temperature measurement. Netsaint is simply the best of the pack.
Oh, a minor thing more; we have never regrettet our small investment in a handheld labeling machine. A small label saying "Cross-over" on a Cat cable or "UPS" on a power cord, saves a lot of trouble.
Todays fiber is not future proof
on
Wiring A New House?
·
· Score: 3, Informative
The question is should I run fiber? I really don't know how much the cable costs since I don't know what cable to use. It is much easier to run cable
before the drywall goes in so I want to make an informed decision now. Ten years from now will I need/want fiber?"
The only reason to run fiber is if the needed cable length, exceed recommended CAT5e length.
Fiber has several disadvantages (I am no cable expert); Connectors; There is a plethora of competing connectors. The most common; SC and ST connectors are rather bulky, and doesn't really work beyond 1Gbps. The new breed of small form connectors for >1Gbps hasn't been certified yet.
If you plan to run fiber "end-to-end", you will have a really limited supply of products to choose from, pay premium prices, and to some extent be tied up with
with whatever connector you choose to begin with.
Changing or making the end-termination on fiber, is much more expensive on fiber, than on copper. (the SMF connector "VF-45" from 3M claims to be the cheapest to use, when doing end-to-end fiber).
The price difference between even a managed 24p 100TX switch and a 24p 100 FX (fiber) switch is significant. The price on 1000 SX fiber equipment is also much higher than 1 GBIC copper equipment.
We just evaluated upgrading our fiber backbone from 10Mbit to either 100FX or 1000SX. (we run a non-profit ISP for 300 apartments), and the price difference between a using a fiber or a copper core switch is very high indeed. If it wasn't for the fact, that our cable length requirements exceed Cat 5e/6, it would probably be cheaper to abandon the fiber and re-wire with copper.
Using fiber-converters (transceivers) also sucks; they cost too much (one needed for every connected device), is yet another source for network problems, and requires yet another power plug.
Actually I believe, that the money saved from investing in fiber and fiber equipment now, could pay for a totally fiber re-wiring if the need for such really should arise into the future.
I really think you will be much better off using copper. People in the know claims, that even though Class D/Cat 6 (200MHz) isn't certified as a standard yet (?), the commercial "Cat 6" cabling systems availably now, should conform to the coming standard. So Cat 6 should be a much better long term investment than fiber. AFAIK Cat 6 should do 10Gbps. Again, people who knows much more than I do, claims that the present standard for multi mode fiber (50 and 62,5 Micron)
doesn't do more than 1,2 Gbps reliably, and is a "dead" standard like Cat 5, and 5e. The coming fiber standards/type (10Gbps) probably named "Laser optimized multimode fiber" LOMMF or OM3 is still in the works.
To summon up;
Cat 5e is the cheap and tried solution. If future needs doesn't go beyond 1Gbps, then why not.
Cat 6 is slightly more expensive, but is much more "future proof" (10Gbps). Perhaps one is still advised to buy a complete system, from the same vendor though.
It allready seems, that Cat 6 rapidly has become the choice when people wire new offices.
Present day fiber standards are not future proof (1,2Gbps), too expensive; not only the fiber cabling systems, but all devices that need to connect with it, and is probably something best left to a professional electrician to install.
Peoples advice about conduit and pull wires, are probably the real key to a long term investment.
A cheap ethernet tester will probably pay for it self too, if you are going to make a lot of cables.
Zeus, Jupiter, and Jove are all the same.. not only in that they're the same god, but also the word itself. The original greek zeus has the stem iov- . in latin, there is no 'J'. the jupiter is literrally
'Father zeus', Iov-pater.. later bastardised into Jupiter.
"Zeus" and "Jupiter" both shares (probably) the same etymology (stems from the same Sanskrit words). Probably refererring to some older indo-european god ("Sky-father".). But "Jupiter" is not a translation, though its meaning is equal, of Zeus, but a independent linguistic development of latin. Best described as independent, parallel development of a common, past god and language.
My point is, that one often reads, that the Latin words for the Roman gods, just are translations of their greek "equivalents". And that is generally not true. (But perhaps a usefull little white lie, when dealing with the profusion of names in greek and roman religion)
While roman religion(s) reimported ideas, and traits from greece, especially after the hellenistic world was subjugated, it would still differ a lot from greek religious practice and "theology".
When I ran windows, Forte Agent from forteinc.com, really was the best nntp client I knew of.
I has some drawbacks though: It was cumbersome to use with more than one news.server, the built-in editor was very simple.
The pros:
Fast to load, fast responsive navigation, good keyboard shortcuts. Self contained in one directory, with nice text based config files, making it very easy to move it around, and preserve settings when reinstalling windows.
Nice, easely used options like; watch or ignore threads. Very powerfull 'per group' options, like.sig settings, whether to download bodies or headers only, expire criterias etc.
I think that Forte Agent was something of a trendsetter regarding GUI newsreaders in its time. And it was and offline newsreader, which really meant something when using a pay per minute modem connection.
Now my preferred nntp client is is KDE Knode, or rather it is not, since our current internet feed provider doesn't have usenet feed:-(, and the free newsservers, don't carry a some of the groups I am interested in.
But so far, Knode seems even more powerfull than Forte Agent in most respects.
Still, it lacks 'per group' expire options (a major letdown for me). And "Score killfiling" is not enough for me, I want real killfilling, eg. delete on the spot.
OTOH, it some really, really nifty features like ">" quote marked text, excellent PGP/GPG support (not sure what to use it for, but I like it anyway)
All in all, Knode is a powerfull nntp client.
Superpimps "Pan" (GNOME) looks nice too (especially if one likes Forte Agent). Haven't used it in a long time, but AFAIK one of its main features, is its abillity to have several, concurrent downloads, from different servers.
Current package management systems in use (rpm, deb, etc.) rely heavily on the package maintainers. You're trusting them on several issues that seem kind of hairy in a large production environment. 1) The binary package does what it's supposed to (read: trojan free)
Eg. Red Hat packages are gpg/pgp signed. A "rpm -K *.rpm" verifies that the package hasn't been tampered with (better securety than md5sum alone).
Besides, all RH packeges comes in source format too (srpm's). So you can manually inspect the source code in a srpm package before building a binary rpm. That way you get the benefit of both "./configure && make && etc" and a package management system.
2) The software within was compiled to an architecture that you can handle (Nothing like finding -i386 meant to your package maintainer that 686 optimizations were included (not so good on some chips, like the AMD k6-2's))
Wouldn't that just mean some kind of (minimal) slowdown? Besides, I would consider "false" optimization a bug, and report it as such.
And what is a K6-2 doing in a "production environment";-)
3) Everything was built with reasonable options
Well, they usually are. But if not (eg. the old ucd-snmp RH rpms, was made without SMUX, and I needed that for my DPT raidcontrollers), then there is always the source rpm. Inside it, is a "spec" file, which among other things, describes what options the package should be compiled with.
Edit that with Vi, build the package, and the problem is solved.
4) The package plays nice and doesn't replace files from other packages on your system.
If so, the installation will fail, complaining about conflicting files. (can be overridden by "--force". old config-files are kept during installation/upgrading, and the new "stock" config files are saved as "package.rpmsave", allowing for manual transition, if the config file format has changed since last release.
To my knowledge "make install" just zaps everything, that gets in its way.
I won't bother with all my other anti-package arguments (dependencies, etc.)I don't get it. Package management systems like rpm are way cool, especially for control freaks:
Lets say I find a mysterious file like "sed" in "/bin". 'rpm -qf sed' tells me what version/package it came from. 'rpm -qi sed' shows the manifest (what does it do). 'rpm -qa | more' shows all packages installed on the system. If I wonder whether I or a script-kiddie has changed some permissions, I can always check them against the cdrom media (tripwire/aide may be better options though).
"dependency" hell are usually caused be the fact, that a needed piece (usually a lib) is missing on the system. A tarball install will fail in that situation too. (if the package is badly packed, eg. you have the lib, but not the package, one can always use "--nodeps" when installing.)
A quick search on www.google.com/linux or rpmfind.net will reveal what package that provides the lib. And rpm will even figure out, in which order, the packages needs be be installed in.
As long as there are people that enjoy slackware, it will keep going. My question to the poster of the article (not that comment I'm replying to) is "When did commercial acceptance become the _only_ thing we care about?"
So true. I now have official reasons to run Linux, but the main reasons I got involved in Linux in the first place, was because it was so much fun, and I still think it is. There is just so many things done right in Linux, and the ability to look behind the "hood" of the system is a joy. Linux is a great text adventure, and even if all commecial distroes disappear, I would still play with it.
Just for the record. I hope (and think) that Slackware doesn't disappear. It was my first real Linux distro, and I like it.
The article points out that they tried to go up against AOL and Yahoo. Might they have lost viewers simply because they're name is too long to bother typing? google.com is easy to type, as is msn.com, etc.
www.av.com (pretty short) takes you to altavista. A whois says the record was created oct. 1998.
No, I dont believe that was the problem.
I did like altavista very much, and unlike google, it did have a good/powerfull "query language".
But they kept loading banners and crap on their frontpage, and the site became very, very sluggish.
And then came google. Google is really what killed altavista: google was extremely fast, and very slick and simple. Their approach to searching really was something new, and they knew how to make money too, without bothering its users ("portals" was the money making hype at that time).
In the beginning, google hadn't indexed so many pages. A search on av, could often give better/more results. But I simply stopped bothering; if it isn't on google or deja.com, I don't really care.
http://cvw.sourceforge.net/ seems to fit your bill somewhat.
The Java client should run on Win, Posix (Linux?) and even Palm (no video I guess).
However, it does require a dedicated document server.
Projects like confman (java too) seems very slick, but is only for people having acces for Mbone (ip6?).
GnomeMeeting (http://www.gnomemeeting.org) is very slick too. It supports the H.323 protocol, so it should work with MS Netmeeting. However no chat, whiteboards etc, only video and voice. (those functions usually requires the T.120 protocol).
So, who can explain to me why this would be worth it? I'm curious to know =) [About water cooling]
Personally I hope not to deal with watercooled client pc's. But such extreme cooling measures might be necessary: A worst case 1.8 P4 consumes aprox. 88 Watts. It looks like future generation CPUs could be even greedier than that (Transmeta is a noticeable exception).
I Imagine that people buying 4-way servers, do so because they expect signicant CPU load for long periods. But 4 x 90/120 Watts (or 8 x 90/120 watts) really generate a lot of heat inside a cabinet. A single future 8 GigaHZ Intel/AMD CPU could also be a challenge for traditionel cooling systems.
However, I find it likely that CPU designers in the future, will sacrifice some speed gains, in order to avoid costly coling systems, such as watercoolers.
I work as a volunteer Sys Admin (BOFH) for my apartment block; 300 users, on a 2mbit leased line, so we are a small time ISP of sorts.
Our users are dynamically assigned private IP numbers, so we use NAT on our gateway.
As I see it, any kind of DoS attack on one of our users, will effectively be an attack on our gateway/firewall, and our commonly shared bandwith.
If such an indiscriminate DoS praxis was instigated by the RIAA against us, we would excersise our legal options to retaliate and defend ourself:
Eg. even though such DoS'ing may become legal in the US, it would still be a criminal activity by my countrys laws (Denmark). Since RIAA has presence in Denmark, it may be possible to persecute them.
Also, perhaps such DoS'ing from the US to other countries, may be illegal even by US law, since it is likely to conflict with international law.
And our humble organisation, might just be politically so well connected, that we could make it an EU case. Certainly we could make it a case in our own parlament, since we occasionally negotiate with high level civil servants, regarding various laws for community(?) based ISPs.
A huge amount of all Danish Internet traffic, goes through the so called DIX. So permanent choke points for RIAA IP numbers there, (and on our backbone providers routers), could also be an option.
We would also bitch and complain to RIAAs backbone provider, suggesting that harbouring DoS script kiddies like RIAA, might be a bad buisness idea, that perhaps could mean trouble for the overseas connectivity for the rest of their costumers (filtering on the DIX, RBL-style, peering agreements, perhaps even lawsuits).
In short, if such a law became a reality in the US, I would strongly advise the RIAA, to individually check the national identity of their DoS-targets IP, before commencing any attack.
This kind of phenomenon is not uncommon already - if you want to buy a rare book, then chances are amazon has it. If not, you google it and find that small site that stocks it.
According to a friend of mine (who sometimes need some really obscure books), Barnes & Noble is an even better "metashop". He doesn't even bother finding the smaller shops that resells the the books to B&N, since he wants the convinience, and "securety" of a brand name shop he knows.
In my country (Denmark) antique bookshops has setup a common metashop and search engine, and are making a "killing" of it, each book store often getting 1500 - 4000$ extra sales per month.
The point is, that these mostly are extra sales, since people like me, are presented with editions we didn't even knew existed. (eg. Juvenalis satires in translation)
All that will happen is that some enterprising guy will set up a meta-shop where you go and enter whatever criteria you like (name, genre etc), and it'll go off and search all the record stores out there. If it isn't on X label, it's on Y instead.
But those kinds of metashops, depends wholly on the subcontractors willingness to cooperate (send in/make available up to date prices and inventory, in some kind of common data format etc.)
If a site _don't_ want to be part of a some meta-price-compare scheme, or product search, they can easely sabotage any attempt. So don't expect a working, common search portal for "cartel-multi-corp-music" and "Small-music-labels".
And apropos metashops; normal online cd-shops are metashops in a way, since they mostly have lots of different music labels, even indie labels.
But this new Behemoth would probably not "resell" its online music to other shops (why would they), making even more difficult for the consumer to get a varied selection.
Re:This version doesn't compile
on
MySQL 4.0 Released
·
· Score: 3, Insightful
GCC is *not* the culprit. Code that doesn't compile on GCC 3 is broken code, that violates the basics of the C language.
True, and I knew that you knew that too, so I did not write it. But a lot of code still doesn't work with gcc 3.x. So as another pointed out, a bug report to the mysql crew is in order. (if you have the time).
I tried to do exactly your commands... same result.
Linux 2.4.12, GCC 3.0.1, SusE 7.0.
No expert at all, but it is likely that GCC 3.0.1 is the culprit. A lot of code seems to choke on version 3.x.x. Eg. RH 7.2beta ships with gcc 2.96 as default (gcc 3.x as an option).
Re:New Sets != Death of Imagination
on
Battle Over Blocks
·
· Score: 5, Insightful
Again, I see yet another adult decrying that the new (more than just rectangles) sets are the death of creativity for kids.
I remember the same complaints in the late 70's. And, I am ashamed to admit, I too, was once complaining about that.
I guess that ones own "golden Lego moments" are frozen with the sets and bricks available at the time. What comes after, seems like follies, and crass commercialism.
I do think that Lego is expensive, especially because the rule about "quantity is a quality in itself", is so true about Lego. Lots of bricks is lots of fun. There is also a certain "critical mass of bricks" needed for many continouse hours of zen-like Lego constuction and play time.
On the other hand, I also think that stuff like Lego are really great toys, far superior to so much else. If for nothing else, because Lego pieces tend to very tough (oh, all the "glittering" plastic trash toys I used to own, who could not withstand even a low intensity afternoon war in my bedroom;)
Sure, roughly 4 nanoseconds after getting it home (only because we banned doing it in the backseat) he has it open and is building it according to the directions -- BUT in a couple of hours he'll have it apart and he'll NEVER build it that way again.
He, he, the frantic art of backseat assembly of Lego sets.
A woman in the one of the articles, is worried because her son only assembles the kits and never take them apart. She blames ready made sets for destroying creativity. But in my childhood (early to mid 70's) when Lego sets were much more simple (and therefore "better"), I knew kids, who would _glue_ the assembled sets together; The horror!
So I think that the "build once, then atomize" or "neatly build, then display" strategies, has much more to do, with the childs basic personality (and age), than with what kinds of sets Lego offers.
Slashdot Mirror
You may be right..but Shouldn't it be 3 people? I mean, how does it help that 2 people have differing numbers? You still need to spend time to figure out which is right. Now if you had three, the two that had the similar input should be the correct one. The occurence of all three being different is significantly less than 1 person being wrong.
There were no explanation in the book. But I think that in those dark days, where "data entry" was a fact, and a profession, that a lot of hard data was gathered about errors in data entries. I don't have the book anymore, but believe that a single person would enter data with around 88% accuracy. The number seems low, but most entry data was probably long (for the typist) meaningless sequences of digits, like Fortran code, scientific measurements etc.
I guess that two people would enter data, a third person would compare data, and recheck the original source if any difference was found. This would allegedly produce a 97-98% accuracy. One could probably improve this somewhat by adding a third data entry person, but it would increase costs, even if one got rid of the "rechecker", and automatically selected the "two of three must right" data, since the rechecker would only have to check the discrepencies, not the whole data set.
So all in all, it is probably a cost-benefit analasys, that decides what one should do. (in the case of this story, I guess a IBM service contract would be the best:-)
I worked with a historian, doing a statistical analasys of approx. 25.000 pages of historical, criminal records.
We only took notes of a few things in each record (type of crime, date, age, sentence etc), but it was gruelling work to do.
I had hacked up a data entry form, but didn't make too much out of the data validation. But after examining the data harvest for the first few days, I decided that real anal data validation was needed, since all kinds of data was wrongly typed, or entered in the wrong fields.
It was not because we were sloppy and we actually took an interrest in the job, but hours and hours of turning pages, gleaning information, type and tab just does make you prone to make mistakes.
I read in some old statistical handbook, that in order to gain 97-98% correct data entries, _two_ people should enter the same data.
I guess that this large retyping of 25.000 tax forms, where each form only seems to be typed in and checked by one person, will be full of errors. I wonder how many thousand record that will needed to be redone.
As I understand it, there are enviromental nasty stuff in several of the components that constitute a typical pc. So pc's aren't so easely, and inexpensive disposed of.
So there is a problem.
A short digression:
Almost all enviromental pollution are caused because, those that pollute finds it cheaper to dump their pollution where others must pay the bill, usually taxpayers or property owners, or just live with the consequenses.
Another factor regarding pollution, and its cleaning up, is, that it is cheap to pollute, but expensive to clean up. And since most pollution is done with a economic motive, those that pollute, rarely pays the bill.
So I believe, that economic incentives, like enviromental fees, is a good way to reduce pollution (which may benefit us all), is fair (since it should payed by those causing the pollution, not by the public at large), and most importantly; may cause a shift from manufacturing and consuming enviromental harmfull stuff, to, at least, lesser harmfull stuff.
Back to the topic:
So if such a fee should be imposed on pc's, it should be done in such a way, that it "hurt" components, that are the most envorimental damageing (eg. cadmium etc), but not those who aren't.
Lets suppose CRT's are enviromental nasties, so lets "fee" them. But if some new CRT technology is not, remember not to leverage a fee on that.
Or else there will be no economic incentive to change behaviour. It is important to keep that goal in sight.
On most motherboards, there sits a tiny battery. I really don't know whether it is enviromental damaging or not, but lets assume it is, but a slightly more expensive battery technology is not.
One can shurely predict, that all motherboard manufactures will use the polluting but cheap solution.
But if the polluting solution becomes more expensive by fees, all the manufactures will turn on a dime, and start using the more enviromental friendly, and cheaper solution.
Eg. At least older Compaq pc's had ATA-IDE cables, that were way shorter than the industri standard.
Alledgely this was because, that when somebody really produces something in bulk, it pays of to save even a few yuans per produced unit.
An email containing a username and password is sent to the registrant a few hours after submission.
As so many others have mentionend, a "few hours" is a very long time. Perhaps not only because people have a short attention span, but because people, the potential costumers, are comparison shopping, and the delay meant that they went elsewere.
I don't know what software your company is making, but lets assume it is something for the desktop user, a piece of software that may have advanteges over the competition, but nevertheless may be easely substituted with something from a competing company. Eg. a small photo editing app.
Somewhere, a proud owner of a new digital camera, wants that kind of software; he goes searching on the web, looks at screenshoots, featurelists and prices, and decides that 3 products looks promising: two of the products are instantly downloaded and tried out, but your app, requires not only a long registration formular, but induces a surpricing two hour waiting period, in which your potential costumer, not only have tried your competitions apps, but may have actually bought them.
Think about going down "Main Street", shopping for a pair of shoes; In one shop, when asking for trying out a pair of shoes, the expedient hands you a two page formular, asking among other things, your phone number, age and job status.
After spending 20 minutes filling it out, you are then told to come back in a couple of hours.
You then go elsewhere.
Lets assume, that your software is somewhat more expensive, and not an "impulse" buy. Perhaps an unique app, that will help people design better, and faster "foo". Surely, professionals may be more patient. But no, that work dead afternoon, where you potential costumer is searching the web for tools that may make him more productive, may be followed up by 5 hectic days. So if you don't engage your potential costumer when he has time, you can loose an oppertunity.
Same thing with the trial period; if your software cost serious money, it probably requieres several hours to test. Most professionals have way too little time at their disposal, they may only have some short timeslots availeably during a week, for testing something new. 30 days may pass quickly, so bump the trial period to 60 days (like eg. IBM does).
In short, make your product as easely availably as possible.
I sure wish they'd figure out how to issue binary diffs instead of complete rpm packages. How much bandwidth was wasted having millions of people download a dozen full packages for the 10 lines of screwed up code in zlib? (No, I don't want to compile it from source. I just want binary packages signed with the disto's gpg key.)
Actually Red Hat does have a binary diff program, called rhmask. Red Hat also used it around RH 5.X/6.0), by issuing "mask" rpms for larger packages like Netscape Communicator.
Even though such diffs seems smart, they trade ease of use, simplicity and less work, for a smaller download size.
eg.: to use a binary diff, you also need access to the original package, which you likely don't have lying around on your server. Then you need to manually generate a new package, and first then, are you able to apply the security patch. More steps to go through, more info to read, more work to do. While the penny-wise modem user may accept that, the overworked server admins don't like it.
So I believe that the rhmask scheme was killed by popular demand.
What about the Red Hat Network? I subscribe 'cause it makes my job as admin SOOOO much easier - but the RHN largely consists of servers with BIG, FAT PIPES.
/pgp signed (+md5sum). So an altered rpm-file, will fail the Up2date agents gpg check against RH's public key, and AFAIK therefore not be automatically installed.
/SPAM /skript-kiddie problem is, that so many boxes, are running totally unattended and / or administrated by less than entirely competent people, who may not even read mail adressed to eg. postmaster@example.com, or without the skills to comprehend the described problem.
As I understand it, it won't be easy to sneak a trojan into RHN, just by cracking RH's RHN servers, since all the RPMS are gpg
So installing Red Hat rpm's, even from "untrusted" mirrors, should be safe, provided that RH can keep their private key secret, and you actually verify the gpg signing (easy to do).
Seems like this also might be an excellent point from which to launch a big DDOS attack, no? How closely does RH watch their servers?
I do think that RH actually monitors their servers (and bandwith), but perhaps much more important, I am quite sure, that they will respond quickly, if people reported a DDoS attack from their IP-range.
IMHO a big part of the DDoS
A 2 hour DoS is nuisance, a 10 day DoS a disaster.
> I probably won't be prioritizing government traffic on any of my routers.
/features on their routers, unless they really had to. /firewalls /sites still doesn't understand the ECN bit, you can lock yourself out, from perhaps 8-10% (*) of the net, if you enable it.
Definitely his/her perogative. But it's such a common sentiment (not trusting the government) I wonder...at what point *would* you voluntarily help out the government to a good end?
When I read the above quotation, I thought the poster meant, that none of his routers are Internet backbone routers, and therefore the proposals are very unlikely to affect his equipment, since prioritized goverment traffic won't pass through them anyway.
Anyway, most sysadmins would probably be very veary about implementing and open up, any new protocols
Eg. ECN (congestion control) is probably a good idea, but since so many routers
(* quote from some site, doesn't know whether they area good estimate or not)
IMHO, the your box is underspecced (ram, ide harddisks) forv m.txt /proc parameters (eg. /proc/sys/vm/ overcommit_memory should be 0 (zero))
p tip2.html
the job you are doing.
Of course, you can try read:
/usr/src/linux[name]/Documentation/sysctl/
for some tunable
Since you are using ide disks, 'man hdparm' is your friend.
Check your kernel config for dma support of your mobo chipset.
Daniel Robbins (from gentoo linux) has written an interesting
article "Maximum swappage" http://www-106.ibm.com/developerworks/library/swa
Linux allow you to parallelize swap, just like a RAID 0 stripe
/etc/fstab:
/dev/hda2 none swap sw,pri=1 0 0
/dev/hdb2 none swap sw,pri=3 0 0
/dev/hdc2 none swap sw,pri=3 0 0
Eg.: spread your swapfile on two disks, with equal priority.
That way, you should in theory, double RW access speed for the
swap. Also, some gains could be gained, if the swap partitions
were moved from disks, that the OS and apps writes to.
But read the article.
A lot of people say "Oh, so DU ammo is dangerous, 'snicker'".
When I was a kid here in Denmark, it was a fairly common event, that some 40 year old WWII "horned" mine was seen drifting into a harbour. I remember that a couple of kids that died, because that rusty old tincan they kicked, in fact was a german stick granade. (those "potatomasher" granades are higly unstable).
Even today it is very common, that fishermen gets a stack of corroded gas granades, usually mustard gas, in their nets, since incredible amounts of WWII gas ammmo, was dumped into the baltic sea after the war.
Some years ago, I visited a woodclearing where german small arms and AA ammo was tried destroyed. It wasn't a well done job; the entire clearing was littered with shells. The holes where the detonations had taken place, was still, after 55 years, without a trace of a single leave of vegetation. Probably caused by the phosphor from the tracer rounds.
The danish coastline was part of the Atlantik Wall, and therefore heavely mined (more than 1.4 million mines). Roughly 1 mineclearer died, for evey ten miles of coastlines, and there are still areas not cleared to this day.
The rest of Europe and the former USSR is littered with WWI and WWII ammo.
The "war most be fought with all means" proponents, really lives in the "here and now", and forget the decades, and centuries that comes afterward, and the huge amount of civilians who has to live on or near the former battlegrounds.
The Soviet Union did not embrace information technology. It was a society that was essentially living in the 1930's. Heavy industry was the priority of the USSR, not semiconductors.
;-).
If you looked on the desks of Soviet desk jockeys in the late 80's, you'd find most offices to be non-computerized (like many western offices). The ones with computers had green screens, IBM or Apple clones. Engineers had Intergraph or Apolla stuff.
The USSR was indeed behind behind the west regarding advanced semiconductor technologi, but your anectdotical evidence can be misleading, since the USSR soviet economy was sharply devided into a civilian part (who got almost nothing) and a military who had first priority.
So even though the standard USSR office was pen-and-paper, the military complex would have access much more advanced technology.
IMHO, soviet military equipment since WWII to until the eighties, was often on par, if not better, than US equipment (especially missilies, tanks, infantery weapons, airplanes, though perhaps not avionics).
OTOH, civilian USSR equipment was always decades behind, what could be found in the west.
The truth isn't bigoted or ignorant.
I believe that a famous USSR newspaper was called "Pravda", meaning "The Truth"
RedHat hates KDE, that's why they do as few packages as possible.
/2.9x source code).
Come on! KDE has been a part of RH distroes since RH 5.2 (?)
Bero (maintainer of the RH KDE rpms) from RH is working very hard on KDE.
He even has a home page with automatically build rpms from daily cvs snapshots.
Their testing style is different, though, and probably more agressive:
KDE 3.0 is found in Red Hats Raw Hide dir. But instead of releasing a whole KDE 3.0 beta 1, then a beta 2, etc, they release batches of incremental improved rpms, often from daily KDE 3 cvs snapshots.
So generaly speaking, RH KDE 3 rpms are much more recent, than those builds found from other vendors on ftp.kde.org.
The upside from this strategy is, that eg. a showstopper bug reported to bugzilla, can be eradicated quickly. The downside is, that a new batch of eg, KDE-networking rpms may break everything, and that everything goes so fast . (eg . I reported a bug known from the changelog, because of that).
One of the reasons, I believe, for RH to sport such an aggressive testing strategy is, that the next RH release will be build entirely with the GCC 3.x
compiler. KDE 3.0 (alpha and first beta at least) couldn't be build with gcc 3.x, and a lot of fixes were probably necessary.
If KDE 3.0 is stable when RH 8.0* hits the market, you can be sure they will include it.
* I think they are going to bump up the version, since an entire distro build around gcc 3.x, is a major step. (compat gcc for all the 2.7x,
And yes! cups (www.cups.org) look like it is going in too. A very, very cool printing system. IBM's JFS is going in too, and of course, KDE 3.0
Regards
How come there's no mention of 'Terminate', wich was by far the best terminal emulation software for DOS? I still keep all my porn on c:\ter400\download\pictures
Terminate is still around, now in a 5.0 version (www.terminate.com).
I don't claim to have tried a lot of terminal emulation programs, but Terminate was really outstanding.
Heat /p4 servers), the equipment can probably survive without active room cooling. Internal cooling of the cabinets may have to be beefed up, especially multi harddisk systems (cheap to do though).
/at jobs.
/router at night, is probably the safest firewall you can get;-)
Unless the room is broom-closet sized, or you got a lot of equipment (more than 5 or 6 Athlon
Perhaps some creativity may help too. Perhaps some of the systems doesn't need to run 24/7.
Some BIOS's have an internal timer and calender, so you can shutdown the systems when likely not in use.
WoL (Wake on LAN) to remote boot, suspend or shutdown systems can be nice too (almost all nics and mobos support WoL nowadays).
Hook it up with some X10 gadgets and a sensor, so that the system(s) boot, if you go near your bedroom console at night, or you alarm clock goes of in the morning, or if you start your coffe machine after 2 o'clock in the night, or...
Other power management features may be present in the OS, so you can suspend the entire system, or just the harddisks, by a cron
Not only will you save some money, but the room will run cooler too.
And unless you run your own DNS, mailserver, etc, then a shut down firewall
Dust
This is my nemesis at the moment, our server room is in a basement, with an untreated cement floor.
I suspect our DAT and some other stuff, died because of the cement dust (ok, so DATs always break down after a short while, but..). Anyway, fans and PSU's seems like dustmagnets, which again leads to worse internal component cooling, so a clean room, without carpets is my recommendation.
Noise
All your equipment will make an infernal noise, and a generally bad indoor clima in the room. Of course, people have very individual sensitivity to this, but personally I prefer to hack outside the serverroom.
I final note, if you run a Linux box, then I can only recommend netsaint, from www.netsaint.org.
It is a very flexible, very reliable monitoring system. Since it checks services with plugins, it is easely extensible to include eg. room temperature measurement. Netsaint is simply the best of the pack.
Oh, a minor thing more; we have never regrettet our small investment in a handheld labeling machine. A small label saying "Cross-over" on a Cat cable or "UPS" on a power cord, saves a lot of trouble.
The question is should I run fiber? I really don't know how much the cable costs since I don't know what cable to use. It is much easier to run cable
/type (10Gbps) probably named "Laser optimized multimode fiber" LOMMF or OM3 is still in the works.
before the drywall goes in so I want to make an informed decision now. Ten years from now will I need/want fiber?"
The only reason to run fiber is if the needed cable length, exceed recommended CAT5e length.
Fiber has several disadvantages (I am no cable expert); Connectors; There is a plethora of competing connectors. The most common; SC and ST connectors are rather bulky, and doesn't really work beyond 1Gbps. The new breed of small form connectors for >1Gbps hasn't been certified yet.
If you plan to run fiber "end-to-end", you will have a really limited supply of products to choose from, pay premium prices, and to some extent be tied up with
with whatever connector you choose to begin with.
Changing or making the end-termination on fiber, is much more expensive on fiber, than on copper. (the SMF connector "VF-45" from 3M claims to be the cheapest to use, when doing end-to-end fiber).
The price difference between even a managed 24p 100TX switch and a 24p 100 FX (fiber) switch is significant. The price on 1000 SX fiber equipment is also much higher than 1 GBIC copper equipment.
We just evaluated upgrading our fiber backbone from 10Mbit to either 100FX or 1000SX. (we run a non-profit ISP for 300 apartments), and the price difference between a using a fiber or a copper core switch is very high indeed. If it wasn't for the fact, that our cable length requirements exceed Cat 5e/6, it would probably be cheaper to abandon the fiber and re-wire with copper.
Using fiber-converters (transceivers) also sucks; they cost too much (one needed for every connected device), is yet another source for network problems, and requires yet another power plug.
Actually I believe, that the money saved from investing in fiber and fiber equipment now, could pay for a totally fiber re-wiring if the need for such really should arise into the future.
I really think you will be much better off using copper. People in the know claims, that even though Class D/Cat 6 (200MHz) isn't certified as a standard yet (?), the commercial "Cat 6" cabling systems availably now, should conform to the coming standard. So Cat 6 should be a much better long term investment than fiber. AFAIK Cat 6 should do 10Gbps. Again, people who knows much more than I do, claims that the present standard for multi mode fiber (50 and 62,5 Micron)
doesn't do more than 1,2 Gbps reliably, and is a "dead" standard like Cat 5, and 5e. The coming fiber standards
To summon up;
Cat 5e is the cheap and tried solution. If future needs doesn't go beyond 1Gbps, then why not.
Cat 6 is slightly more expensive, but is much more "future proof" (10Gbps). Perhaps one is still advised to buy a complete system, from the same vendor though.
It allready seems, that Cat 6 rapidly has become the choice when people wire new offices.
Present day fiber standards are not future proof (1,2Gbps), too expensive; not only the fiber cabling systems, but all devices that need to connect with it, and is probably something best left to a professional electrician to install.
Peoples advice about conduit and pull wires, are probably the real key to a long term investment.
A cheap ethernet tester will probably pay for it self too, if you are going to make a lot of cables.
Zeus, Jupiter, and Jove are all the same.. not only in that they're the same god, but also the word itself. The original greek zeus has the stem iov- . in latin, there is no 'J'. the jupiter is literrally
'Father zeus', Iov-pater.. later bastardised into Jupiter.
"Zeus" and "Jupiter" both shares (probably) the same etymology (stems from the same Sanskrit words). Probably refererring to some older indo-european god ("Sky-father".). But "Jupiter" is not a translation, though its meaning is equal, of Zeus, but a independent linguistic development of latin. Best described as independent, parallel development of a common, past god and language.
My point is, that one often reads, that the Latin words for the Roman gods, just are translations of their greek "equivalents". And that is generally not true. (But perhaps a usefull little white lie, when dealing with the profusion of names in greek and roman religion)
While roman religion(s) reimported ideas, and traits from greece, especially after the hellenistic world was subjugated, it would still differ a lot from greek religious practice and "theology".
When I ran windows, Forte Agent from forteinc.com, really was the best nntp client I knew of. .sig settings, whether to download bodies or headers only, expire criterias etc.
/GPG support (not sure what to use it for, but I like it anyway)
I has some drawbacks though: It was cumbersome to use with more than one news.server, the built-in editor was very simple.
The pros:
Fast to load, fast responsive navigation, good keyboard shortcuts. Self contained in one directory, with nice text based config files, making it very easy to move it around, and preserve settings when reinstalling windows.
Nice, easely used options like; watch or ignore threads. Very powerfull 'per group' options, like
I think that Forte Agent was something of a trendsetter regarding GUI newsreaders in its time. And it was and offline newsreader, which really meant something when using a pay per minute modem connection.
Now my preferred nntp client is is KDE Knode, or rather it is not, since our current internet feed provider doesn't have usenet feed:-(, and the free newsservers, don't carry a some of the groups I am interested in.
But so far, Knode seems even more powerfull than Forte Agent in most respects.
Still, it lacks 'per group' expire options (a major letdown for me). And "Score killfiling" is not enough for me, I want real killfilling, eg. delete on the spot.
OTOH, it some really, really nifty features like ">" quote marked text, excellent PGP
All in all, Knode is a powerfull nntp client.
Superpimps "Pan" (GNOME) looks nice too (especially if one likes Forte Agent). Haven't used it in a long time, but AFAIK one of its main features, is its abillity to have several, concurrent downloads, from different servers.
Current package management systems in use (rpm, deb, etc.) rely heavily on the package maintainers. You're trusting them on several issues that seem kind of hairy in a large production environment.
;-)
/upgrading, and the new "stock" config files are saved as "package.rpmsave", allowing for manual transition, if the config file format has changed since last release.
/package it came from. 'rpm -qi sed' shows the manifest (what does it do). 'rpm -qa | more' shows all packages installed on the system. If I wonder whether I or a script-kiddie has changed some permissions, I can always check them against the cdrom media (tripwire /aide may be better options though).
1) The binary package does what it's supposed to (read: trojan free)
Eg. Red Hat packages are gpg/pgp signed. A "rpm -K *.rpm" verifies that the package hasn't been tampered with (better securety than md5sum alone).
Besides, all RH packeges comes in source format too (srpm's). So you can manually inspect the source code in a srpm package before building a binary rpm. That way you get the benefit of both "./configure && make && etc" and a package management system.
2) The software within was compiled to an architecture that you can handle (Nothing like finding -i386 meant to your package maintainer that 686 optimizations were included (not so good on some chips, like the AMD k6-2's))
Wouldn't that just mean some kind of (minimal) slowdown? Besides, I would consider "false" optimization a bug, and report it as such.
And what is a K6-2 doing in a "production environment"
3) Everything was built with reasonable options
Well, they usually are. But if not (eg. the old ucd-snmp RH rpms, was made without SMUX, and I needed that for my DPT raidcontrollers), then there is always the source rpm. Inside it, is a "spec" file, which among other things, describes what options the package should be compiled with.
Edit that with Vi, build the package, and the problem is solved.
4) The package plays nice and doesn't replace files from other packages on your system.
If so, the installation will fail, complaining about conflicting files. (can be overridden by "--force". old config-files are kept during installation
To my knowledge "make install" just zaps everything, that gets in its way.
I won't bother with all my other anti-package arguments (dependencies, etc.)I don't get it. Package management systems like rpm are way cool, especially for control freaks:
Lets say I find a mysterious file like "sed" in "/bin". 'rpm -qf sed' tells me what version
"dependency" hell are usually caused be the fact, that a needed piece (usually a lib) is missing on the system. A tarball install will fail in that situation too. (if the package is badly packed, eg. you have the lib, but not the package, one can always use "--nodeps" when installing.)
A quick search on www.google.com/linux or rpmfind.net will reveal what package that provides the lib. And rpm will even figure out, in which order, the packages needs be be installed in.
As long as there are people that enjoy slackware, it will keep going. My question to the poster of the article (not that comment I'm replying to) is "When did commercial acceptance become the _only_ thing we care about?"
So true. I now have official reasons to run Linux, but the main reasons I got involved in Linux in the first place, was because it was so much fun, and I still think it is. There is just so many things done right in Linux, and the ability to look behind the "hood" of the system is a joy. Linux is a great text adventure, and even if all commecial distroes disappear, I would still play with it.
Just for the record. I hope (and think) that Slackware doesn't disappear. It was my first real Linux distro, and I like it.
The article points out that they tried to go up against AOL and Yahoo. Might they have lost viewers simply because they're name is too long to bother typing? google.com is easy to type, as is msn.com, etc.
/powerfull "query language".
/more results. But I simply stopped bothering; if it isn't on google or deja.com, I don't really care.
www.av.com (pretty short) takes you to altavista. A whois says the record was created oct. 1998.
No, I dont believe that was the problem.
I did like altavista very much, and unlike google, it did have a good
But they kept loading banners and crap on their frontpage, and the site became very, very sluggish.
And then came google. Google is really what killed altavista: google was extremely fast, and very slick and simple. Their approach to searching really was something new, and they knew how to make money too, without bothering its users ("portals" was the money making hype at that time).
In the beginning, google hadn't indexed so many pages. A search on av, could often give better
http://cvw.sourceforge.net/ seems to fit your bill somewhat.
The Java client should run on Win, Posix (Linux?) and even Palm (no video I guess).
However, it does require a dedicated document server.
Projects like confman (java too) seems very slick, but is only for people having acces for Mbone (ip6?).
GnomeMeeting (http://www.gnomemeeting.org) is very slick too. It supports the H.323 protocol, so it should work with MS Netmeeting. However no chat, whiteboards etc, only video and voice. (those functions usually requires the T.120 protocol).
So, who can explain to me why this would be worth it? I'm curious to know =) [About water cooling]
/120 Watts (or 8 x 90 /120 watts) really generate a lot of heat inside a cabinet. A single future 8 GigaHZ Intel /AMD CPU could also be a challenge for traditionel cooling systems.
Personally I hope not to deal with watercooled client pc's. But such extreme cooling measures might be necessary: A worst case 1.8 P4 consumes aprox. 88 Watts. It looks like future generation CPUs could be even greedier than that (Transmeta is a noticeable exception).
I Imagine that people buying 4-way servers, do so because they expect signicant CPU load for long periods. But 4 x 90
However, I find it likely that CPU designers in the future, will sacrifice some speed gains, in order to avoid costly coling systems, such as watercoolers.
I work as a volunteer Sys Admin (BOFH) for my apartment block; 300 users, on a 2mbit leased line, so we are a small time ISP of sorts. /firewall, and our commonly shared bandwith.
Our users are dynamically assigned private IP numbers, so we use NAT on our gateway.
As I see it, any kind of DoS attack on one of our users, will effectively be an attack on our gateway
If such an indiscriminate DoS praxis was instigated by the RIAA against us, we would excersise our legal options to retaliate and defend ourself:
Eg. even though such DoS'ing may become legal in the US, it would still be a criminal activity by my countrys laws (Denmark). Since RIAA has presence in Denmark, it may be possible to persecute them.
Also, perhaps such DoS'ing from the US to other countries, may be illegal even by US law, since it is likely to conflict with international law.
And our humble organisation, might just be politically so well connected, that we could make it an EU case. Certainly we could make it a case in our own parlament, since we occasionally negotiate with high level civil servants, regarding various laws for community(?) based ISPs.
A huge amount of all Danish Internet traffic, goes through the so called DIX. So permanent choke points for RIAA IP numbers there, (and on our backbone providers routers), could also be an option.
We would also bitch and complain to RIAAs backbone provider, suggesting that harbouring DoS script kiddies like RIAA, might be a bad buisness idea, that perhaps could mean trouble for the overseas connectivity for the rest of their costumers (filtering on the DIX, RBL-style, peering agreements, perhaps even lawsuits).
In short, if such a law became a reality in the US, I would strongly advise the RIAA, to individually check the national identity of their DoS-targets IP, before commencing any attack.
This kind of phenomenon is not uncommon already - if you want to buy a rare book, then chances are amazon has it. If not, you google it and find that small site that stocks it.
/make available up to date prices and inventory, in some kind of common data format etc.)
According to a friend of mine (who sometimes need some really obscure books), Barnes & Noble is an even better "metashop". He doesn't even bother finding the smaller shops that resells the the books to B&N, since he wants the convinience, and "securety" of a brand name shop he knows.
In my country (Denmark) antique bookshops has setup a common metashop and search engine, and are making a "killing" of it, each book store often getting 1500 - 4000$ extra sales per month.
The point is, that these mostly are extra sales, since people like me, are presented with editions we didn't even knew existed. (eg. Juvenalis satires in translation)
All that will happen is that some enterprising guy will set up a meta-shop where you go and enter whatever criteria you like (name, genre etc), and it'll go off and search all the record stores out there. If it isn't on X label, it's on Y instead.
But those kinds of metashops, depends wholly on the subcontractors willingness to cooperate (send in
If a site _don't_ want to be part of a some meta-price-compare scheme, or product search, they can easely sabotage any attempt. So don't expect a working, common search portal for "cartel-multi-corp-music" and "Small-music-labels".
And apropos metashops; normal online cd-shops are metashops in a way, since they mostly have lots of different music labels, even indie labels.
But this new Behemoth would probably not "resell" its online music to other shops (why would they), making even more difficult for the consumer to get a varied selection.
GCC is *not* the culprit. Code that doesn't compile on GCC 3 is broken code, that violates the basics of the C language.
True, and I knew that you knew that too, so I did not write it. But a lot of code still doesn't work with gcc 3.x. So as another pointed out, a bug report to the mysql crew is in order. (if you have the time).
I tried to do exactly your commands... same result.
Linux 2.4.12, GCC 3.0.1, SusE 7.0.
No expert at all, but it is likely that GCC 3.0.1 is the culprit. A lot of code seems to choke on version 3.x.x. Eg. RH 7.2beta ships with gcc 2.96 as default (gcc 3.x as an option).
Again, I see yet another adult decrying that the new (more than just rectangles) sets are the death of creativity for kids.
I remember the same complaints in the late 70's. And, I am ashamed to admit, I too, was once complaining about that.
I guess that ones own "golden Lego moments" are frozen with the sets and bricks available at the time. What comes after, seems like follies, and crass commercialism.
I do think that Lego is expensive, especially because the rule about "quantity is a quality in itself", is so true about Lego. Lots of bricks is lots of fun. There is also a certain "critical mass of bricks" needed for many continouse hours of zen-like Lego constuction and play time.
On the other hand, I also think that stuff like Lego are really great toys, far superior to so much else. If for nothing else, because Lego pieces tend to very tough (oh, all the "glittering" plastic trash toys I used to own, who could not withstand even a low intensity afternoon war in my bedroom;)
Sure, roughly 4 nanoseconds after getting it home (only because we banned doing it in the backseat) he has it open and is building it according to the directions -- BUT in a couple of hours he'll have it apart and he'll NEVER build it that way again.
He, he, the frantic art of backseat assembly of Lego sets.
A woman in the one of the articles, is worried because her son only assembles the kits and never take them apart. She blames ready made sets for destroying creativity. But in my childhood (early to mid 70's) when Lego sets were much more simple (and therefore "better"), I knew kids, who would _glue_ the assembled sets together; The horror!
So I think that the "build once, then atomize" or "neatly build, then display" strategies, has much more to do, with the childs basic personality (and age), than with what kinds of sets Lego offers.