I too run Redhat, but have suffered
Annoyances.
I'm contemplating going over to SuSE since some of our colleagues at IBM use SuSE as well.
However there are some issues you may wish to consider. Redhat has some very talented developers (e.g. Alan Cox), and has a fair number of developers who develop drivers (e.g. the Adaptec dptio drivers), some of which I use. However I use SuSE at home and it has become popular within our organization. While SuSE has a few annoyances of its own, I'll need to test it on our hardware to evaluate its fitness for use.
I would suggest getting a boxed set of Suse Professional (its like $80 U.S.) and playing with it for an afternoon or 2 on your hardware. SuSE comes with many useful packages bundled that Redhat won't carry (for political reasons?).
Some SuSE technical information is conveyed in German language news groups, if you can read a bit, it does help (often avoids the "how can I...?" kind of questions). SuSE's upgrade system works pretty well.
Actually Albany went completely dead, my neighborhood went back online at 8:30 or so last night. Our governor (George Pataki) was making noises like the problem might have been in Canada, but perhaps people will remember to keep asking him questions (he has a tremendous teflon persona, and frequently just smiles waves and walks away if he doesn't want to answer questions).
I'm not sure of numbers, a quick google turned
up U.S. Metropolitan Area Population Estimates, you can check their
ranked list. I'm not sure how they draw the boundaries, for example, there are people in eastern PA who commute to NYC and people in Albany NY who also commute to NYC (it is 2 1/2 to 3 hours by car, in Albany you can use a train, it is about the same distance). Many major western hemisphere cities were omitted (Mexico City in particular is huge, and Canada has some major cities too).
I would really like a high quality well accepted Linux/Unix based accounting package. However, most programmers I know don't find writing accounting packages "fun", so they don't work on them except at gunpoint (or for a paycheck).
Thus, I think the ideal solution would be for the project team to generate revenue, either by support or find a paying customer (who would allow release of the source). Suppose they wrote a book and released a free CD of the source code with it?
Would that generate enough royalties? This may be hard in the current economic climate, but I think it would give them their best chances. Would vendors who are making big Linux pushes be interested? Have the project leaders directly solicited input (and contributions) from these vendors (e.g. IBM)?
Thanks. Going to see if anyone locally carries them, and get one ASAP.
I've seen them at radio shack (but I couldn't see it in the on-line catalog. I seem to recall the price being around $30 or so, but the on-line prices appeared a bit higher for other vendors so I was conservative in my original post, because I didn't trust my memory on the price).
The best way to measure is probably experimentally.
You could use a "Kill a Watt" device, that has
a combination of a male and female plug with an output for indicating how much power is being drawn.
You can get one for $40 U.S. or less.
Actually, I had heard that linuxconf had problems, but I never actually heard what problems lead to it being dropped? Was there some deep seated problem(s)? Why did they drop it?
Breaking the install so that an upgrade hosed my Athlon box at home (motivating a quick run to Best Buy to get SuSE, and I've never looked back).
Tip: Select another kernel from the boot menu or insert the boot floppy you made before.
Redhat detected the Athlon and installed a customized kernel. Since I was at home, I didn't have another box on hand to surf the web and find the work around for this. The really beautiful thing was that this was the point at which RH introduced ext3 support, and I decided to upgrade all my ext2 partitions to ext3, so I COULD NEITHER ROLL BACK NOR SEARCH ONLINE FOR A FIX. The distro picked the wrong kernel and kernel parameters and stuck it on my HD, not me. Eventually I upgraded, to a different distro. But don't worry, after I reported the Bug to Redhat, they suggested that I try the install again and set the noathlon option in the kernel
boot parameters. That was beautiful, like I was going to remove a working distro install off my now stable and happy home machine to give them another chance.
Distros just don't do proper integration testing
on
Worst Linux Annoyances?
·
· Score: 4, Informative
I've got a lab, and we rolled out redhat due to popularity and have stuck with it since 1998.
Since then, Redhat has been suprisingly sloppy in their distributions, and I'm just about ready to drop them for another distro (maybe SuSE).
Among my beefs (these occurred in different versions) are:
Inconsistency in the administration tools,
including dropping the linuxconf tool for the less functional controlpanel.
Failure to include any updates to Netscape.
Choosing an immature unrealeased beta gcc version for a production release.
Breaking the NFS client so that acccess times
became 100X slower (way to go guys, great job not testing there!).
Breaking the install so that an upgrade hosed my Athlon box at home (motivating a quick run to Best Buy to get SuSE, and I've never looked back).
Numerous Kernel bugs induced during "upgrades" which I need to accept to close security holes. I had 6 months of hell due to a Kernel bug which caused my server to give up the ghost without a cry for help. Sure I blamed it on hardware at first, since I had 1 year of uptime, but then I realized that their updates just didn't cut it, and they finally fixed it this June.
SuSE has some glitches too, in particular
My X server leaks memory (allegedly due to Anti Aliasing of fonts), so I have to close my X windows and restart it every few weeks.
SuSE doesn't properly listen for the hostname my ISP assigns so ssh can't set the display variable correctly when remoting in.
Many of the installed games don't start up when I select them from the menu.
The drivers for the video card sometimes hang when my daugther plays tux racer.
Exactly what I thought. I imagine things like inlining and other compiler optimisations might confuse things further.
I looked through the talk slides in pdf, and they discuss some of the approaches.
I think the tool works, but it has some limits.
From what I can tell, the tool disassembles a binary and identifies the basic blocks (the longest pieces of code that must run from start to finish, so they start on a labeled statement or after a branch/jump/call instruction and end before a labeled statement or branch/jump/call instruction), and establishes a control graph, indicating which basic blocks invoke which other basic blocks. Some types of known unsafe instructions can be identified, and the tool back tracks through the basic blocks looking to see how inputs are manipulated to these instructions, and what the triggers might be. Some graph analysis tricks are done, and they can back track up to 64 levels (I don't know if they mean instructions or basic blocks). This kind of tool sounds very useful to me. Consider the following scenarios:
Suppose you worked for a software customer has large exposure in the event of a security compromise (e.g. stock market/bank/medical data/social security numbers/military data). They want to limit their risk exposure. They can now check binaries for some known vulnerabilities before they are deployed and make detailed reports to their vendors. They may be able to cancel contracts with vendors who deliver code that is shown to be unsafe, or they may make passing these tests a precondition to acceptance, or they could work with the vendor and release detailed reports to them so they can fix the problems.
Imagine you are a software vendor, and you have the source code to your project. You can usually compile extra symbol table information in
for debugging, so that you can map data and instruction locations back to source code.
I do agree that source code analysis tools may be informative, but if you have cross language development or have to link to libraries that you don't have the source for, this sort of tool could be really helpful.
On a side note, when I develop software, I find
many bugs are actually in error handlers,
since they are seldom executed and not heavily tested. Being able to analyze code that is not
ordinarily executed is very useful.
You could offer firewall installation/configuration
and teach them how to protect their data.
Wireless setup/lan cards could be nice too
(along with some end-to-end encryption tools
for wireless, or a more functional link layer encryption tool).
frankly I'd like something to come out on linux or a BSD that has to have windows drivers created, rather then vica-versa, something so bold, inventive, and paradigm shifting that it can't be ignored
Have you ever heard of the BSD TCP/IP protocol stack? That might qualify (except MS used the BSD code as a starting point).
Do you still have to shut down and hand edit a configuration file to set the user agent? One thing that KDE and opera do well is allow you to specify what user agent string your browser will send a particular web site. This is probably the only reason I don't use galeon or Mozilla (although I have off the shelf builds from SUSE 8.0 at home).
The on-line manuals are helpful except when you are either stuck during installation/configuration or have network problems and you are trying to troubleshoot. Then hard copy is indispensible.
Its not just money, wealthy folks sometimes run as 3rd party candidates (Ross Perot for President, Tom Golisano for NY state Governor). Sure they sometimes do goofy things (in particular Perot), but often times they just can't get votes because voters prefer the 2 party system.
Well, it wasn't intended to imply their demise. Rather I meant to say that while they are still pursuing selling on portables/mobiles/cell phones, this is actually a well established business approach of theirs (they've been at it for years).
Although the market is very soft, going for
installation on hand held devices (cell phone/PDA) devices is likely to give them a leading share. Since the vendors seemed gun shy of microsoft, I'd give the Opera folks a bigger chance if that market should materialize in a reasonable time frame. The big concern is if Microsoft can just buy their way in, they may squeeze the little guys out.
Evidently the editors missed their
own original posting of this story.
It might be a good idea for editors to adopt a
policy of browsing the recent stories
(1-2 days old) to verify they don't
duplicate (editors should not degrade content).
The materials required are just one expense,
the catalyst typically is expected to be reusable
(consumed at a very low rate due to inefficiencies).
However, the amount of raw material required to
extract the energy, the size of the apparatus and
the amount of energy required to get a unit of
energy are probably the real issues. If it takes
more than a Joule to extract the hydrogen required
to generate a Joule of energy, the system is only
viable for special applications, not as an energy
source.
Interestingly, one of the QNX founders is named Gordon Bell, which is coincidentally the name of a very famous system architect at Digital and later at Microsoft Research. Finally this article gave enough information I was able to convince myself that it was just a name space collision, they aren't the same fellow.
Can you hold this over your managers head for compensation during the next performance review? It is worth a shot to mention it to him/her in clear language- I am a team player. I am busting hump. I want this reflected in my performance evaluations.
I'd suggest trying to negotiate a
performance review mid project.
If you wait until the project is
done the employer may give weak reasons to avoid
giving you a raise. Charging annoyance fees
is sensible if you can live with the annoyance.
Additionally, depending on your home life,
you may wish to negotiate for flex time and the ability to work from home (e.g. if you have kids, how will they be picked up/dropped off from daycare). You may also wish to ask for vacation guarantees and extra salary (or a bonus if this is a one time thing) for successful completion, with rewards built in for quick completion and quality work. You may also ask the senior management point blank if this is the way they plan to do business in the future.
Actually faculty go to conferences to learn about new things, where they attend lectures. A good presentation often means that I'll read the paper,
and schmoozing with good people is helpful in
getting new ideas and improving my understanding.
Most interesting developments in C.S. occur at conferences (due to their quick review cycles).
I've been surprised by students in the past who
ask questions that I don't expect. Sometimes students open topics.
Many Professors feel that they have some sort of special message that can only be transmitted in a lecture.
Actually lecture is not my favorite way of teaching, but when there are many students and
only one of instructor, it will have to do. Regarding text books and teaching,
the instructor has a responsibility to
pick what they think is important and not
to merely find a textbook and parrot it.
Now for entry level courses, there
are often adequate text books, but for upper
division courses, really good textbooks are hard
to find. I cannot for example find a really
good book on Operating Systems. Other fields
like Compiler design (the Dragon Book),
programming languages (Michael Scott's book),
automata theory (Aho Hopcroft and Ullman),
Algorithms (Coreman Lieserson and Rivest)
and computer architecture (Hennessy and
Patterson) all have classic books.
Computer Networking has improved books,
but still has a way to go.
I don't have time (yet) to write my own,
but some day I might.
Slashdot Mirror
However there are some issues you may wish to consider. Redhat has some very talented developers (e.g. Alan Cox), and has a fair number of developers who develop drivers (e.g. the Adaptec dptio drivers), some of which I use. However I use SuSE at home and it has become popular within our organization. While SuSE has a few annoyances of its own, I'll need to test it on our hardware to evaluate its fitness for use.
I would suggest getting a boxed set of Suse Professional (its like $80 U.S.) and playing with it for an afternoon or 2 on your hardware. SuSE comes with many useful packages bundled that Redhat won't carry (for political reasons?). Some SuSE technical information is conveyed in German language news groups, if you can read a bit, it does help (often avoids the "how can I...?" kind of questions). SuSE's upgrade system works pretty well.
Actually Albany went completely dead, my neighborhood went back online at 8:30 or so last night. Our governor (George Pataki) was making noises like the problem might have been in Canada, but perhaps people will remember to keep asking him questions (he has a tremendous teflon persona, and frequently just smiles waves and walks away if he doesn't want to answer questions).
I'm not sure of numbers, a quick google turned up U.S. Metropolitan Area Population Estimates, you can check their ranked list. I'm not sure how they draw the boundaries, for example, there are people in eastern PA who commute to NYC and people in Albany NY who also commute to NYC (it is 2 1/2 to 3 hours by car, in Albany you can use a train, it is about the same distance). Many major western hemisphere cities were omitted (Mexico City in particular is huge, and Canada has some major cities too).
Thus, I think the ideal solution would be for the project team to generate revenue, either by support or find a paying customer (who would allow release of the source). Suppose they wrote a book and released a free CD of the source code with it? Would that generate enough royalties? This may be hard in the current economic climate, but I think it would give them their best chances. Would vendors who are making big Linux pushes be interested? Have the project leaders directly solicited input (and contributions) from these vendors (e.g. IBM)?
The best way to measure is probably experimentally. You could use a "Kill a Watt" device, that has a combination of a male and female plug with an output for indicating how much power is being drawn. You can get one for $40 U.S. or less.
Actually, I had heard that linuxconf had problems, but I never actually heard what problems lead to it being dropped? Was there some deep seated problem(s)? Why did they drop it?
- Inconsistency in the administration tools,
including dropping the linuxconf tool for the less functional controlpanel.
- Failure to include any updates to Netscape.
- Choosing an immature unrealeased beta gcc version for a production release.
- Breaking the NFS client so that acccess times
became 100X slower (way to go guys, great job not testing there!).
- Breaking the install so that an upgrade hosed my Athlon box at home (motivating a quick run to Best Buy to get SuSE, and I've never looked back).
- Numerous Kernel bugs induced during "upgrades" which I need to accept to close security holes. I had 6 months of hell due to a Kernel bug which caused my server to give up the ghost without a cry for help. Sure I blamed it on hardware at first, since I had 1 year of uptime, but then I realized that their updates just didn't cut it, and they finally fixed it this June.
SuSE has some glitches too, in particular- Suppose you worked for a software customer has large exposure in the event of a security compromise (e.g. stock market/bank/medical data/social security numbers/military data). They want to limit their risk exposure. They can now check binaries for some known vulnerabilities before they are deployed and make detailed reports to their vendors. They may be able to cancel contracts with vendors who deliver code that is shown to be unsafe, or they may make passing these tests a precondition to acceptance, or they could work with the vendor and release detailed reports to them so they can fix the problems.
- Imagine you are a software vendor, and you have the source code to your project. You can usually compile extra symbol table information in
for debugging, so that you can map data and instruction locations back to source code.
On a side note, when I develop software, I find many bugs are actually in error handlers, since they are seldom executed and not heavily tested. Being able to analyze code that is not ordinarily executed is very useful.I do agree that source code analysis tools may be informative, but if you have cross language development or have to link to libraries that you don't have the source for, this sort of tool could be really helpful.
Wireless setup/lan cards could be nice too (along with some end-to-end encryption tools for wireless, or a more functional link layer encryption tool).
Do you still have to shut down and hand edit a configuration file to set the user agent? One thing that KDE and opera do well is allow you to specify what user agent string your browser will send a particular web site. This is probably the only reason I don't use galeon or Mozilla (although I have off the shelf builds from SUSE 8.0 at home).
The on-line manuals are helpful except when you are either stuck during installation/configuration or have network problems and you are trying to troubleshoot. Then hard copy is indispensible.
Its not just money, wealthy folks sometimes run as 3rd party candidates (Ross Perot for President, Tom Golisano for NY state Governor). Sure they sometimes do goofy things (in particular Perot), but often times they just can't get votes because voters prefer the 2 party system.
Although the market is very soft, going for installation on hand held devices (cell phone/PDA) devices is likely to give them a leading share. Since the vendors seemed gun shy of microsoft, I'd give the Opera folks a bigger chance if that market should materialize in a reasonable time frame. The big concern is if Microsoft can just buy their way in, they may squeeze the little guys out.
But if she got all your stuff, how smart are you really a genius?
Evidently the editors missed their own original posting of this story. It might be a good idea for editors to adopt a policy of browsing the recent stories (1-2 days old) to verify they don't duplicate (editors should not degrade content).
Does Estonia have a strong engineering/science tradition? Is the university system particluarly strong there?
The materials required are just one expense, the catalyst typically is expected to be reusable (consumed at a very low rate due to inefficiencies). However, the amount of raw material required to extract the energy, the size of the apparatus and the amount of energy required to get a unit of energy are probably the real issues. If it takes more than a Joule to extract the hydrogen required to generate a Joule of energy, the system is only viable for special applications, not as an energy source.
Interestingly, one of the QNX founders is named Gordon Bell, which is coincidentally the name of a very famous system architect at Digital and later at Microsoft Research. Finally this article gave enough information I was able to convince myself that it was just a name space collision, they aren't the same fellow.
Additionally, depending on your home life, you may wish to negotiate for flex time and the ability to work from home (e.g. if you have kids, how will they be picked up/dropped off from daycare). You may also wish to ask for vacation guarantees and extra salary (or a bonus if this is a one time thing) for successful completion, with rewards built in for quick completion and quality work. You may also ask the senior management point blank if this is the way they plan to do business in the future.
Actually faculty go to conferences to learn about new things, where they attend lectures. A good presentation often means that I'll read the paper, and schmoozing with good people is helpful in getting new ideas and improving my understanding. Most interesting developments in C.S. occur at conferences (due to their quick review cycles). I've been surprised by students in the past who ask questions that I don't expect. Sometimes students open topics.