The first being, you cannot run an all NT shop.
The second is similar, you cannot run an all
Unix shop.
That's rubbish.
I've worked in a large organization that migrated to all NT 4.0 (4500 desktops, large WAN infrastructure). The only non-NT OS's in house were "appliance" stuff like IOS for the routers and Cisco's Pix's OS for the firewall. We used IOS for the same reason we didn't use FreeBSD or Solaris to route packets - routers are for routing, not a general purpose operating system.
I've also worked in places (admittedly small numbers of desktops, but large numbers of customers) where everything was Linux based. Because they didn't need any fancy calendering features (they saw each other every day) and they didn't do any documentation (eventually they did HTML doco), Linux was the right choice.
In most common-law countries, it is a requirement that you obey the law, even if you're unaware of the details of the law. There is extensive case history that people do not get away with the "I did not know that I had to do that" excuse.
In this case, both the owners (SOCOG) of the Olympics site and IBM are responsible for making the site comply with all applicable local laws, such as it may not contain R rated pictures without an AVS. If there were such a section and there wasn't an AVS, then the owners would be required to take it down or fit one within a short time (usually the period of the take down notice) to make it comply with local laws.
The problem is that in the building industry, builders, architects and engineers are all required to comply with strict building codes. In the wild west of the Internet, there are not too many laws and the laws that do exist it is unclear who has the duty to implement them.
For example, I have to advise my customer (a large telco) on various matters. I keep upto date with various things, and I give them my best advice. When it comes to things like implementing LEA interception devices, my ethics gets in the way of me implementing said devices even though that may cost me work. But it is a legal requirement to have them in the place I work at. This is the counter alternative to what happened with the Olympics web site. I'm sure there are decent people at IBM who raised the issue, and were hounded down by project managers with an eye on scope creep. You can't blame IBM for trying to eliminate scope creep, but accessibility is a legal requirement, and should have been on the original contract. If this were the building industry, IBM would have been duty bound to advise their customer that they need to include x,y,z provisions in order to comply with the law. Then IBM could have had that negotiated into the contract and earnt additional dollars from implementing it. Now, it's just a no win situation for all concerned.
You obviously haven't worked with a big outsourcer before. This is standard fare. I was recently quoted over $500k to make a small three story building move from a single shared collision domain 10 Mbs hubs to a switched 100 Mb/s network for approx 80 nodes. Direct hardware costs were about $20k, and a bit of network admin time to swap the stuff in and out as everyone already had 100 Mb/s capable NICs and the wiring was all CAT5.
Two million is cheap for something the outsourcer obviously does not want to do (it's extra-scope, therefore maintenance fees apply, each change costs $$$$, where each change is likely to be rejigging a single image as well more costly items like rejigging the backend to support ALT tags).
Money for jam.
I've read all the >2 posts and I think most of you are missing the point. The problem is not that James Murdoch saying that Mandarin usage is going to take over the net, but simply that most programs, OS's and web front ends simply cannot handle most non-English languages very well.
Try this as an exercise. Insert some Japanese or double byte language into your favorite web site, such as slashdot. It wont work. The text at the bottom of my post says "Sayonara". Not hard, and I'm even using an encoding that fits in UTF-8. But I bet myself $0.05 that it will not show up correctly and may even stuff the rest of the posts. I managed to mangle the linux-kernel mail list archives at kernel.org. Bring on language DoS!
Now, you probably can't read Japanese, but why should code break because I don't use an ASCII encoding?
If you're interested in real productivity, you must try two different environments that work:
Offices with a door, no flourescent lighting and most importantly a single occupant
The well connected home office
Read Peopleware by Demarco and Lister for the reasoning and proof of why this is so.
I've had the pleasure of working in both of these environments, and I'm easily 20-30 times more productive than when I'm in a cubicle environment (like I am at the moment). I know this because I managed to write 30 pages of documentation in two days whereas it took 50 days to write the first 20 pages in my most recent contract. I've repeated this experiment many times over the last two years and hope that you will take the time to test it too.
I'm looking forward to the Olympics so I can telecommute again and get some real productivity back from the black hole of time that is my "office" at my current multinational telco client.
I've worked in ISP's before. The abuse queue at one of them (a very large one) jumped from an undercurrent of about 1000 outstanding items to over 1200 on the basis of one single spam incident. The ISP costs each abuse incident requiring action at $25 to reply and fix. Thus this one spam incident cost the ISP more than $5000 to manage and resolve, and that doesn't take into account the good will aspects. The bandwidth stolen from the ISP and the customer costs money as well, and to maintain a responsive system, most tier 1 ISP's have excess capacity. Spam is not really a big consumer of bandwidth unless you happen to be the sucker with an open relay, but the management costs are astronomical. In addition, of the twenty or so times I saw the results of abusive "customers" who bought $20 pre-paid internet accounts and injected several million messages per account before having it closed, the account costs this ISP many hundreds of dollars. The headers are all forged (who do you want to be today?), the recipients entirely unwilling. The mail administrators in one of the worst incidents worked until 2 am fixing this problem up. Does the spammer pay for this? I don't think so. If the local mail relays are full of unwanted messages from non-paying (or abusive) customers they cannot service the other 100,000+ customers legitimate traffic in a timely fashion. If they paid *all* the intervening ISPs (as if) for the full cost of their actions, and everyone opted in rather than the other way around, this would not be a major problem. It's not about free speech, but simply this: A is stopped from sending to C,D,E...n by B. A is stealing from C,D,E,..n's ISP and from C,D,E,..n, and from many intervening networks, and thus many managers and administrators do no like this loss, thus signing up for B's service. "A" does not pay for the privilege and they forge their identity. Why does "A" think they have a *right* to steal? UCE is theft of service. End of story.
The reason you'd be killed is not because your car weighs less or has the windows at the wrong height (although they are factors), it's because the car is so rigid that you'd be like an egg in a toolbox.
It really aggravates me to see more and more SUV's on the road. It's not too long before driving cars like my Toyota Echo (Yaris in most parts of the world, about 850 kg dry) is dangerous not because it is one the safest small cars in the world, but because of this single equation:
p = mv
SUV's steal my safety by being heavier, more structurally rigid (due to marketing requirements, not safety requirements) and usually driven by aggressive drivers who completely overestimate their own driving skills. The only solution is a move towards more tank-like cars, which imposes a heavy burden on our environment.
My best wishes to Cassie, and to Jason - I don't know you, but I hope you get well soon.
It was Tim Norman who wrote pbm2ppa, and I'm one of the lead developers (for the last few months) of the color replacement, pnm2ppa.
It's ironic that they chose to use SourceForge, where we are also located. See this for our project.
It would be nice if HP released the specs for their PPA printers, but we have received their final word on it, so we'll just have to reverse engineer the protocol more fully and hope to $DEITY that we don't kill anyone's printer.
But I see the HP thing in a good light. Printing under Unix sucks. End of story. There needs to be proper bi-directional communication in a standard way on all Unixes, proper paper handling, easy to install drivers, easy to use multi-paper size print jobs and the ability to control the resolution on the fly. Until these things are fixed^Wwritten, printing on Unix will always be a third class citizen.
In the MS world, the most potent and prevalent virii are macro viruses. Over time, MS built in limited forms of checks to prevent the average user from doing bad stuff to themselves, although the problem is still bad.
The KOffice team are about to release a product that has corba access to all other components and is scriptable using python. Python has no security model and can import many useful modules to do pretty much anything you could do via C or C++ to your system. If a Koffice script can look up your address book via the ORB and run system(), Melissa/koffice is well on the way to being written - and very successful with those running Koffice.
Linux is not immune to virii, and never will be. Most distributions are out to win the tick box war, which means packages+++. I can assure you that no one has the time to check all the packages being installed; and some of them are huge. So bugs will be there, which creates exploit opportunities. Just one common get root exploit is enough to allow a virii to propogate easily.
A smug smart arse attitude to Windows users will not help once the virus kits start getting distributed. Take the problem seriously, and use proper precautions:
In the rest of the world, GSM rules because their local legislature made it so - as the sole standard for their country. I can call anywhere in Australia and many other contries without changing my handset or the SIM card in it - I just turn it on and it works. I can drive from Melbourne to Cairns (about 4000 km by road) and have full service along practically the entire route with my 100 gm GSM handset. I can change carriers but retain my handset if I wanted to. I can take my SIM card (about 2 grams in weight) with me and use a friend's phone in Singapore or Scotland if that tickled my fancy.
In the US, you can barely use some mobile phones within your own city, let alone another state. CDMA, TDMA, AMPS, PCS, and a little GSM thrown in; you name an half assed protocol, and the US has it. It's bad for consumers.
Another example, already seen in this discussion is AM stereo. In Australia, FM Stereo was made mandatory with one single high quality format. It has been so successful, that today no one listens to AM for music. The government made a lot of money from AM music stations willing to move to FM to retain their listeners.
Then the govt experimented and was burnt with self regulation, and the result was AM stereo. Complicated by two competing standards, consumers didn't buy the sets and car manufacturers settled on CQAM, but it was too late. I am not aware of any stations still transmitting in AM stereo after less than 15 years after its introduction.
As a consumer, don't be afraid of a single standard. Be afraid of a single carrier, a single cable company, a single software company.
I was once like you; I always had other things to do with my time. Eventually it got to a be hard habit. Now, at 29, I'm stuck in my ways.
I don't go out (much) - and I don't like to go out, which is the important factor for me. The places I like to go out to are basically selfish (ie films, restaurants, and concerts).
I'm still single even though my chances of meeting the right person are doubled (I'm bi).
I was watching an old French film last night, Cousin Cousine, and one of the lead characters said somehting I feel is profound: "Always do one unpredictable thing every day". I think that's important. So what's my message for all you young 'uns? Get out there IRL and meet people.
with the advent of reparse points and letterless drives in Windows 2000 (this was possible under NT 4.0, but not used), this is not really an advantage that's going to last long. I estimate a late December-early January for Win2K to be in the stores.
Text based log files are not remotely administrable, so you must establish some form of secure local administration practice such as ssh to manage them. That doesn't scale if you have to manage (say) 1,000 servers.
The list should be seen as a list of things to do by the Linux developers. Not as a thing to respond directly to. You'll never compete with MS's marketing people. So don't.
There's no point. His staff have a standing order to dump mail on this matter. His e-mail is filtered to drop any reference to this matter from outside.gov.au sites.
In my immediate past site (I work at various sites), we had two Exchange bridgeheads distributing large numbers of tasks to 54 sites. Over a million messages a week being sent around the state to manage custom developed software. Beyond hardware failure (can't blame the OS for that one), this worked without a hitch.
Exchange doesn't BSOD with large number of messages. Exchange is generally not the cause of BSOD's - NT might do so for other reasons (bad RAM, dodgy hardware, the occasional bug check), but I've never seen one caused by Exchange. I've seen Exchange being very sick and AV all over the place, but it doesn't cause NT to crash. It's like saying that sendmail can oops a Linux kernel. Doesn't happen.
Look through the following two articles for the current cumulative bug fix list for Exchange 5.5 SP3. If anyone can find a single referenced Q article that mentions Exchange code causing a BSOD, I'll donate $AUD20 to Amnesty International. This is not about plain crashes, this about BSOD, which are different beasties altogether.
Moderation sucking again. Re:Exchange of Opinions
on
CNN on Sendmail for NT
·
· Score: 1
I normally read with threshold:2 to avoid crappy articles, and I wandered down to threshold:0 to see what I have been missing, particularly on obviously potential anti-MS articles such as this.
What I see is a bias in the moderation against MS informational and "from the field" articles. Most of the useful articles in this thread should have been 2, informational. Instead they are 1 or 0.
I know that/. is basically a Linux-against-the-world, a beastary of geeks type of place. And that will be reflected not only in moderation, but also meta-moderation. The problem is that we need to have a better balance, like an "experts in the field" super moderators, who can only bounce articles up to 2 (say) that are truly informational and relevant. To become a super moderator, you would have to apply, and get recognition from people to say that you are one of:
wrote the original code
administrate lots of users in interesting places
I am an active developer of that tool
I am Linus
I'm not just saying this for MS products, I'm saying for everything, whether you happen to be the FreeBSD package master for a product, or a XFree86 developer (like I was), or a kernel developer. We need peer review by qualified individuals. This is not about moderating up trollish opinions, this about moderating up informational articles of any flavour.
To me, I get value from/. by giving me the devil's advocate view of MS. I work with MS products all the time - I secure and administrate them in very large sites (I'm currently working in one of Australia's largest IT sites with over 30,000 desktops directly affected by my work). I have to have a reality check from time to time, and/. is the right place for that. (rant) With the exception of the week I had off after the truly appalling Richard Stevens incident - I almost didn't come back. Who needs a community that disrepectful of someone who IMHO constitutes a stellar light in a 99.9% dark universe? He did so much for Unix, and my faith in humanity so troubled by/.'s response. (rant finished)
As you go through life (I'm all of 28, and probably one of the older people here), one of things you should learn is that you have to take into account information from both sides, knowing that some or all of that information is flawed and biased. Then you munge the data and make up your own mind. It'll still come out something like "M$ sux", but hell, you've had a think, and you've made up your own mind. Don't let anyone else do that for you.
The US chose a crap standard. One that doesn't allow mobile reception (IMHO not such a bad thing especially when cars are involved), forces lower quality (ie stations are able to not use HD images, and therefore use their allocation for more channels of lower quality images...). The US standard also doesnt have progressive scan (1080i means interlaced).
It's the like the GSM vs US hodge podge (CDMA, GSM, Satellite, AMPS, etc, etc, etc) again. Political porkbarrelling matters much more than the consumer. Again.
Compatibility - there's a major amount of computing power in the set. They can have a analog and digital tuner to allow existing signals to be processed. If you want digitial reception for old equipment, I'm sure someone will build a set top box to convert the digital signal to NTSC, like when UHF became more common.
A large TV I'm looking at has a slot for a digital tuner card. Since it already does 100 Hz progressive scan, it'll cope with HD Digital TV as going to be found in Europe and Australia.
I was going to ask this as well - but in a slightly more broad terms.
Alan,
When are we going to see a multi-threaded I/O subsystem/framework, where non-brain dead devices can be utilised to their maxiumum potential on SMP machines like mine?
When I set up a production environment (regardless of operating system chosen), the first step is always to have policy and change control.
Change Control
You must have change control or you will suffer downtime. Downtime represents a transaction rate of 0 trans/sec, which is clearly unacceptable.
Development and Acceptance Test
You must have a separate machine for development, and another machine for acceptance test. Of the two machines, only the A.T. machine must be identical to your production server. Otherwise A.T. simply cannot replicate the environment you're going to test, and thus any testing is at best misleading, or at worst, completely invalid.
You must create a set of repeatable build instructions that takes you from a fresh blank machine to a stable, reliable, working production system. And you should have a set of tests that thoroughly gives the resulting systems a complete workout, including sustained load, boundary condiditions (such as empty rows), and attacks against the system whilst trying to continue to process transactions.
Finally, the best advice I can give you is don't skimp on reliability and availability. Buy RAID with hot rebuild. Buy a server with redundant PSU's, and not a handmade machine. Buy an additional NIC per machine, and put that on a different switch - dual path everything.
In terms of SQL and web based stuff, from a security standpoint, it's always advisable to have your SQL server behind a firewall (or at least on a separate private network).
In terms of speed, I've always found that having enough RAM to allow several outer joins to complete in RAM really helps. As someone else mentioned, it's a good idea to index columns you select on a regular basis.
Make sure you can dump the database online - stopping the dbms whilst a dump takes place is unacceptable; if it takes 30 minutes, that's reduced your availability from near 100% to 97.9%. That's bad.
Yep, I spose I found the code a confusing at first. But there is the documentation area on mozilla.org that guides you through the major sections, and there is the news groups where you can ask the people who either wrote it in the first place or look after it now.
If you're doing the Linux front end, you're going to be interested in less than 10-20% of the code, so it's not too bad. If you're profiling, you might visit a particular.c file, and you'll end up knowing a lot about that.c file. Most times it's not truly necessary to know everything about the area you might be working with. For example, I tried making a new Timer.cpp to replace the current Unix one (which relies, wrongly on the existance of X11). I got to know those few files very well, until I basically gave up.
The documentation is reasonable, particularly for NGLayout (or whatever its called this month).
Re:Will Mozilla ever gain outside contributors?
on
Mozilla M9 Released
·
· Score: 2
Yes - I was at one stage contributing actively to the Rhapsody/MacOS X Server port. I got out of that when Apple canned the x86 port of MacOS X Server - something that is still 100% boneheaded. My Rhapsody DR2 install no longer works, so basically Apple lost a developer FOREVER. Oh yeah, I was going through a bit of a down patch in my life, and I wasn't coding much. But I did have commit privs on the CVS tree for Mozilla/Rhapsody, so it is possible.
If you have an itch, scratch it - download the source, get cvs in order and sync with the latest source.
Even if you can't code, get the latest stuff, build it (it's not hard) and run it through its paces. If you're going to debug the lizard, you'll need mucho memory (on my 96 MB dual PPro, it just swapped...). But documenters, bug testers and more than just the occasional fix really help. High quality bug reports are worth their weight in gold.
Maybe think about profiling. If memory or CPU usage is bugging you, compile with the profiler options turned on and figure out where the problem is, and fix it (or suggest a fix) based upon your research.
I was watching the great BBC series The Planets last night on ABC Australia, and it was the episode on the Sun. This is great series btw, very approachable.
The interviewees described what they think was the heliopause interacting with a coronal mass ejection. Voyager 2 monitored the CME pass at +100 days from the Sun, and then it monitored a massive magnetic disturbance at +400 days. So NASA thinks that the heliopause caused this disturbance, and it's approximately 4 times the distance from sun as Voyager is now, which means that Voyager will meet the heliopause in another ~ 60-65 years (uncertainty on my part - Voyager accelerated every time it bypassed a planet, but hasn't since Neptune), as it has been travelling for 22 years now.
I don't know if the RTGs or Voyager itself will survive that long.
RC3 is due by Oct 7, I doubt that the release code will be ready on Oct 7 timeline. RC2 is Aug 25. I'm feeling for a November RTM, meaning in the shops for Christmas. We'll know more once RC2 is out and happening. I have a number of bugs I need chased up before I'd personally be happy with the release.
Xfree86 3.9 when I was working on it (from Aug 96 - May 98) used a great deal less memory than XFree86 3.3.x. That's because it only loads the chipset, color and depth support you're currently using. I don't know about memory leaks, but if you're really worried about it, why not join the effort and help out?
XFree86 has been able to support multi-plane overlays for a while. When I was working on the Matrox Millennium driver, starting around August 96, the support got added eventually by one of the other dudes, um I can't remember if is was one of the other Andrews or Radek. It was easily two years ago that support was added - I haven't worked on the driver for a year now.
There's just no easy way to make a different plane depth in X as it was shipped, and only some chipsets (eg, Matrox and a very _few_ others) can support it. I certainly never tested it, although I knew it was there.
Slashdot Mirror
That's rubbish.
I've worked in a large organization that migrated to all NT 4.0 (4500 desktops, large WAN infrastructure). The only non-NT OS's in house were "appliance" stuff like IOS for the routers and Cisco's Pix's OS for the firewall. We used IOS for the same reason we didn't use FreeBSD or Solaris to route packets - routers are for routing, not a general purpose operating system.
I've also worked in places (admittedly small numbers of desktops, but large numbers of customers) where everything was Linux based. Because they didn't need any fancy calendering features (they saw each other every day) and they didn't do any documentation (eventually they did HTML doco), Linux was the right choice.
In this case, both the owners (SOCOG) of the Olympics site and IBM are responsible for making the site comply with all applicable local laws, such as it may not contain R rated pictures without an AVS. If there were such a section and there wasn't an AVS, then the owners would be required to take it down or fit one within a short time (usually the period of the take down notice) to make it comply with local laws.
The problem is that in the building industry, builders, architects and engineers are all required to comply with strict building codes. In the wild west of the Internet, there are not too many laws and the laws that do exist it is unclear who has the duty to implement them.
For example, I have to advise my customer (a large telco) on various matters. I keep upto date with various things, and I give them my best advice. When it comes to things like implementing LEA interception devices, my ethics gets in the way of me implementing said devices even though that may cost me work. But it is a legal requirement to have them in the place I work at. This is the counter alternative to what happened with the Olympics web site. I'm sure there are decent people at IBM who raised the issue, and were hounded down by project managers with an eye on scope creep. You can't blame IBM for trying to eliminate scope creep, but accessibility is a legal requirement, and should have been on the original contract. If this were the building industry, IBM would have been duty bound to advise their customer that they need to include x,y,z provisions in order to comply with the law. Then IBM could have had that negotiated into the contract and earnt additional dollars from implementing it. Now, it's just a no win situation for all concerned.
You obviously haven't worked with a big outsourcer before. This is standard fare. I was recently quoted over $500k to make a small three story building move from a single shared collision domain 10 Mbs hubs to a switched 100 Mb/s network for approx 80 nodes. Direct hardware costs were about $20k, and a bit of network admin time to swap the stuff in and out as everyone already had 100 Mb/s capable NICs and the wiring was all CAT5. Two million is cheap for something the outsourcer obviously does not want to do (it's extra-scope, therefore maintenance fees apply, each change costs $$$$, where each change is likely to be rejigging a single image as well more costly items like rejigging the backend to support ALT tags). Money for jam.
Try this as an exercise. Insert some Japanese or double byte language into your favorite web site, such as slashdot. It wont work. The text at the bottom of my post says "Sayonara". Not hard, and I'm even using an encoding that fits in UTF-8. But I bet myself $0.05 that it will not show up correctly and may even stuff the rest of the posts. I managed to mangle the linux-kernel mail list archives at kernel.org. Bring on language DoS!
Now, you probably can't read Japanese, but why should code break because I don't use an ASCII encoding?
æÈç
Read Peopleware by Demarco and Lister for the reasoning and proof of why this is so.
I've had the pleasure of working in both of these environments, and I'm easily 20-30 times more productive than when I'm in a cubicle environment (like I am at the moment). I know this because I managed to write 30 pages of documentation in two days whereas it took 50 days to write the first 20 pages in my most recent contract. I've repeated this experiment many times over the last two years and hope that you will take the time to test it too.
I'm looking forward to the Olympics so I can telecommute again and get some real productivity back from the black hole of time that is my "office" at my current multinational telco client.
I've worked in ISP's before. The abuse queue at one of them (a very large one) jumped from an undercurrent of about 1000 outstanding items to over 1200 on the basis of one single spam incident. The ISP costs each abuse incident requiring action at $25 to reply and fix. Thus this one spam incident cost the ISP more than $5000 to manage and resolve, and that doesn't take into account the good will aspects. The bandwidth stolen from the ISP and the customer costs money as well, and to maintain a responsive system, most tier 1 ISP's have excess capacity. Spam is not really a big consumer of bandwidth unless you happen to be the sucker with an open relay, but the management costs are astronomical. In addition, of the twenty or so times I saw the results of abusive "customers" who bought $20 pre-paid internet accounts and injected several million messages per account before having it closed, the account costs this ISP many hundreds of dollars. The headers are all forged (who do you want to be today?), the recipients entirely unwilling. The mail administrators in one of the worst incidents worked until 2 am fixing this problem up. Does the spammer pay for this? I don't think so. If the local mail relays are full of unwanted messages from non-paying (or abusive) customers they cannot service the other 100,000+ customers legitimate traffic in a timely fashion. If they paid *all* the intervening ISPs (as if) for the full cost of their actions, and everyone opted in rather than the other way around, this would not be a major problem. It's not about free speech, but simply this: A is stopped from sending to C,D,E...n by B. A is stealing from C,D,E,..n's ISP and from C,D,E,..n, and from many intervening networks, and thus many managers and administrators do no like this loss, thus signing up for B's service. "A" does not pay for the privilege and they forge their identity. Why does "A" think they have a *right* to steal? UCE is theft of service. End of story.
When you're ready to come out, please send in an article to Slashdot. Maybe they'll post it. Until then, please remove yourself from the genepool.
The reason you'd be killed is not because your car weighs less or has the windows at the wrong height (although they are factors), it's because the car is so rigid that you'd be like an egg in a toolbox.
It really aggravates me to see more and more SUV's on the road. It's not too long before driving cars like my Toyota Echo (Yaris in most parts of the world, about 850 kg dry) is dangerous not because it is one the safest small cars in the world, but because of this single equation:
p = mv
SUV's steal my safety by being heavier, more structurally rigid (due to marketing requirements, not safety requirements) and usually driven by aggressive drivers who completely overestimate their own driving skills. The only solution is a move towards more tank-like cars, which imposes a heavy burden on our environment.
My best wishes to Cassie, and to Jason - I don't know you, but I hope you get well soon.
It was Tim Norman who wrote pbm2ppa, and I'm one of the lead developers (for the last few months) of the color replacement, pnm2ppa.
It's ironic that they chose to use SourceForge, where we are also located. See this for our project.
It would be nice if HP released the specs for their PPA printers, but we have received their final word on it, so we'll just have to reverse engineer the protocol more fully and hope to $DEITY that we don't kill anyone's printer.
But I see the HP thing in a good light. Printing under Unix sucks. End of story. There needs to be proper bi-directional communication in a standard way on all Unixes, proper paper handling, easy to install drivers, easy to use multi-paper size print jobs and the ability to control the resolution on the fly. Until these things are fixed^Wwritten, printing on Unix will always be a third class citizen.
In the MS world, the most potent and prevalent virii are macro viruses. Over time, MS built in limited forms of checks to prevent the average user from doing bad stuff to themselves, although the problem is still bad.
The KOffice team are about to release a product that has corba access to all other components and is scriptable using python. Python has no security model and can import many useful modules to do pretty much anything you could do via C or C++ to your system. If a Koffice script can look up your address book via the ORB and run system(), Melissa/koffice is well on the way to being written - and very successful with those running Koffice.
Linux is not immune to virii, and never will be. Most distributions are out to win the tick box war, which means packages+++. I can assure you that no one has the time to check all the packages being installed; and some of them are huge. So bugs will be there, which creates exploit opportunities. Just one common get root exploit is enough to allow a virii to propogate easily.
A smug smart arse attitude to Windows users will not help once the virus kits start getting distributed. Take the problem seriously, and use proper precautions:
In the rest of the world, GSM rules because their local legislature made it so - as the sole standard for their country. I can call anywhere in Australia and many other contries without changing my handset or the SIM card in it - I just turn it on and it works. I can drive from Melbourne to Cairns (about 4000 km by road) and have full service along practically the entire route with my 100 gm GSM handset. I can change carriers but retain my handset if I wanted to. I can take my SIM card (about 2 grams in weight) with me and use a friend's phone in Singapore or Scotland if that tickled my fancy.
In the US, you can barely use some mobile phones within your own city, let alone another state. CDMA, TDMA, AMPS, PCS, and a little GSM thrown in; you name an half assed protocol, and the US has it. It's bad for consumers.
Another example, already seen in this discussion is AM stereo. In Australia, FM Stereo was made mandatory with one single high quality format. It has been so successful, that today no one listens to AM for music. The government made a lot of money from AM music stations willing to move to FM to retain their listeners.
Then the govt experimented and was burnt with self regulation, and the result was AM stereo. Complicated by two competing standards, consumers didn't buy the sets and car manufacturers settled on CQAM, but it was too late. I am not aware of any stations still transmitting in AM stereo after less than 15 years after its introduction.
As a consumer, don't be afraid of a single standard. Be afraid of a single carrier, a single cable company, a single software company.
I don't go out (much) - and I don't like to go out, which is the important factor for me. The places I like to go out to are basically selfish (ie films, restaurants, and concerts).
I'm still single even though my chances of meeting the right person are doubled (I'm bi).
I was watching an old French film last night, Cousin Cousine, and one of the lead characters said somehting I feel is profound: "Always do one unpredictable thing every day". I think that's important. So what's my message for all you young 'uns? Get out there IRL and meet people.
with the advent of reparse points and letterless drives in Windows 2000 (this was possible under NT 4.0, but not used), this is not really an advantage that's going to last long. I estimate a late December-early January for Win2K to be in the stores.
Text based log files are not remotely administrable, so you must establish some form of secure local administration practice such as ssh to manage them. That doesn't scale if you have to manage (say) 1,000 servers.
The list should be seen as a list of things to do by the Linux developers. Not as a thing to respond directly to. You'll never compete with MS's marketing people. So don't.
In my immediate past site (I work at various sites), we had two Exchange bridgeheads distributing large numbers of tasks to 54 sites. Over a million messages a week being sent around the state to manage custom developed software. Beyond hardware failure (can't blame the OS for that one), this worked without a hitch.
Exchange doesn't BSOD with large number of messages. Exchange is generally not the cause of BSOD's - NT might do so for other reasons (bad RAM, dodgy hardware, the occasional bug check), but I've never seen one caused by Exchange. I've seen Exchange being very sick and AV all over the place, but it doesn't cause NT to crash. It's like saying that sendmail can oops a Linux kernel. Doesn't happen.
Look through the following two articles for the current cumulative bug fix list for Exchange 5.5 SP3. If anyone can find a single referenced Q article that mentions Exchange code causing a BSOD, I'll donate $AUD20 to Amnesty International. This is not about plain crashes, this about BSOD, which are different beasties altogether.
Part One
Part Two
What I see is a bias in the moderation against MS informational and "from the field" articles. Most of the useful articles in this thread should have been 2, informational. Instead they are 1 or 0.
I know that /. is basically a Linux-against-the-world, a beastary of geeks type of place. And that will be reflected not only in moderation, but also meta-moderation. The problem is that we need to have a better balance, like an "experts in the field" super moderators, who can only bounce articles up to 2 (say) that are truly informational and relevant. To become a super moderator, you would have to apply, and get recognition from people to say that you are one of:
I'm not just saying this for MS products, I'm saying for everything, whether you happen to be the FreeBSD package master for a product, or a XFree86 developer (like I was), or a kernel developer. We need peer review by qualified individuals. This is not about moderating up trollish opinions, this about moderating up informational articles of any flavour.
To me, I get value from /. by giving me the devil's advocate view of MS. I work with MS products all the time - I secure and administrate them in very large sites (I'm currently working in one of Australia's largest IT sites with over 30,000 desktops directly affected by my work). I have to have a reality check from time to time, and /. is the right place for that. (rant) With the exception of the week I had off after the truly appalling Richard Stevens incident - I almost didn't come back. Who needs a community that disrepectful of someone who IMHO constitutes a stellar light in a 99.9% dark universe? He did so much for Unix, and my faith in humanity so troubled by /.'s response. (rant finished)
As you go through life (I'm all of 28, and probably one of the older people here), one of things you should learn is that you have to take into account information from both sides, knowing that some or all of that information is flawed and biased. Then you munge the data and make up your own mind. It'll still come out something like "M$ sux", but hell, you've had a think, and you've made up your own mind. Don't let anyone else do that for you.
It's the like the GSM vs US hodge podge (CDMA, GSM, Satellite, AMPS, etc, etc, etc) again. Political porkbarrelling matters much more than the consumer. Again.
Compatibility - there's a major amount of computing power in the set. They can have a analog and digital tuner to allow existing signals to be processed. If you want digitial reception for old equipment, I'm sure someone will build a set top box to convert the digital signal to NTSC, like when UHF became more common.
A large TV I'm looking at has a slot for a digital tuner card. Since it already does 100 Hz progressive scan, it'll cope with HD Digital TV as going to be found in Europe and Australia.
I was going to ask this as well - but in a slightly more broad terms.
Alan,
When are we going to see a multi-threaded I/O subsystem/framework, where non-brain dead devices can be utilised to their maxiumum potential on SMP machines like mine?
keep up the good work
When I set up a production environment (regardless of operating system chosen), the first step is always to have policy and change control.
Change Control
You must have change control or you will suffer downtime. Downtime represents a transaction rate of 0 trans/sec, which is clearly unacceptable.
Development and Acceptance Test
You must have a separate machine for development, and another machine for acceptance test. Of the two machines, only the A.T. machine must be identical to your production server. Otherwise A.T. simply cannot replicate the environment you're going to test, and thus any testing is at best misleading, or at worst, completely invalid.
You must create a set of repeatable build instructions that takes you from a fresh blank machine to a stable, reliable, working production system. And you should have a set of tests that thoroughly gives the resulting systems a complete workout, including sustained load, boundary condiditions (such as empty rows), and attacks against the system whilst trying to continue to process transactions.
Finally, the best advice I can give you is don't skimp on reliability and availability. Buy RAID with hot rebuild. Buy a server with redundant PSU's, and not a handmade machine. Buy an additional NIC per machine, and put that on a different switch - dual path everything.
In terms of SQL and web based stuff, from a security standpoint, it's always advisable to have your SQL server behind a firewall (or at least on a separate private network).
In terms of speed, I've always found that having enough RAM to allow several outer joins to complete in RAM really helps. As someone else mentioned, it's a good idea to index columns you select on a regular basis.
Make sure you can dump the database online - stopping the dbms whilst a dump takes place is unacceptable; if it takes 30 minutes, that's reduced your availability from near 100% to 97.9%. That's bad.
Good luck!
Yep, I spose I found the code a confusing at first. But there is the documentation area on mozilla.org that guides you through the major sections, and there is the news groups where you can ask the people who either wrote it in the first place or look after it now.
If you're doing the Linux front end, you're going to be interested in less than 10-20% of the code, so it's not too bad. If you're profiling, you might visit a particular .c file, and you'll end up knowing a lot about that .c file. Most times it's not truly necessary to know everything about the area you might be working with. For example, I tried making a new Timer.cpp to replace the current Unix one (which relies, wrongly on the existance of X11). I got to know those few files very well, until I basically gave up.
The documentation is reasonable, particularly for NGLayout (or whatever its called this month).
Yes - I was at one stage contributing actively to the Rhapsody/MacOS X Server port. I got out of that when Apple canned the x86 port of MacOS X Server - something that is still 100% boneheaded. My Rhapsody DR2 install no longer works, so basically Apple lost a developer FOREVER. Oh yeah, I was going through a bit of a down patch in my life, and I wasn't coding much. But I did have commit privs on the CVS tree for Mozilla/Rhapsody, so it is possible.
If you have an itch, scratch it - download the source, get cvs in order and sync with the latest source.
Even if you can't code, get the latest stuff, build it (it's not hard) and run it through its paces. If you're going to debug the lizard, you'll need mucho memory (on my 96 MB dual PPro, it just swapped...). But documenters, bug testers and more than just the occasional fix really help. High quality bug reports are worth their weight in gold.
Maybe think about profiling. If memory or CPU usage is bugging you, compile with the profiler options turned on and figure out where the problem is, and fix it (or suggest a fix) based upon your research.
I was watching the great BBC series The Planets last night on ABC Australia, and it was the episode on the Sun. This is great series btw, very approachable.
The interviewees described what they think was the heliopause interacting with a coronal mass ejection. Voyager 2 monitored the CME pass at +100 days from the Sun, and then it monitored a massive magnetic disturbance at +400 days. So NASA thinks that the heliopause caused this disturbance, and it's approximately 4 times the distance from sun as Voyager is now, which means that Voyager will meet the heliopause in another ~ 60-65 years (uncertainty on my part - Voyager accelerated every time it bypassed a planet, but hasn't since Neptune), as it has been travelling for 22 years now.
I don't know if the RTGs or Voyager itself will survive that long.
RC3 is due by Oct 7, I doubt that the release code will be ready on Oct 7 timeline. RC2 is Aug 25. I'm feeling for a November RTM, meaning in the shops for Christmas. We'll know more once RC2 is out and happening. I have a number of bugs I need chased up before I'd personally be happy with the release.
Xfree86 3.9 when I was working on it (from Aug 96 - May 98) used a great deal less memory than XFree86 3.3.x. That's because it only loads the chipset, color and depth support you're currently using. I don't know about memory leaks, but if you're really worried about it, why not join the effort and help out?
XFree86 has been able to support multi-plane overlays for a while. When I was working on the Matrox Millennium driver, starting around August 96, the support got added eventually by one of the other dudes, um I can't remember if is was one of the other Andrews or Radek. It was easily two years ago that support was added - I haven't worked on the driver for a year now.
There's just no easy way to make a different plane depth in X as it was shipped, and only some chipsets (eg, Matrox and a very _few_ others) can support it. I certainly never tested it, although I knew it was there.
Andrew