In respect for kernel code, a maintainer of a project is someone who collates the amendments to code, merges the patches together, and forwards them onto Linus or whoever. They also control style and namespace pollution.
Check the/usr/src/linux/MAINTAINERS for a full list.
In the case of my sound card driver, there was no maintainer. As the driver at the time supported at least 4 different chipsets, and I was adding support for another, I had no means to verify that my changes were indeed correct, as I was unable to locate any of the other models of that card. I did have the specs for the other chipsets, but I've been burnt by incorrect specs in the past. Luckily so far no-one has reported that my changes have broken things. But 2.4.0 hasn't gone live yet!
Maintainers are useful when multiple changes are being made to the same code, and the changes are incompatable with each other. There are many active maintainers within Linux, but still vast ranges of code without anyone to oversee it.
There is a lot of unmaintained code in the Linux kernel as well. Device drivers for obsolete hardware, seldom used filesystems, that kind of thing.
Usually this is due to the original author no longer having access to working hardware, or a platform that the hardware will work in.
Last year I tried to investigate why my sound card's Midi port wouldn't work with the drivers in the kernel. My sound card was on a weird daughterboard and not easy to remove without a hacksaw. Once I finally pulled the thing out, I dicovered that the main chips model number was later than any of those listed in the driver.
An email to the Linux Kernel mailing list received one reply, from Alan Cox, stating that no-one was now maintaining that code. In the end I amended the code myself after reverse engineering a DOS device driver. It didn't help that no specs for this particular chip were available.
Simillary the UMSDOS filesystem was broken in the 2.1.x kernel code for over 50 revisions, as the addition of dentries broke it big style. Once someone else took over the code things started to work, but it wasn't really reliable until very recently.
Read the story of Robin Hood and Friar Tuck in the Jargon File. I don't have it to hand at present, so some details may be incorrect.
This relates to a system around 25 years ago, which had a severe security problem, which the vendors refused to fix. The security hole affected the monitor process, and made it possible to patch into it, IIRC.
Some hackers, disgruntled over the laxness of security, decided to exploit the problem on one of the manufacturers own systems. This installed two processes, called Robin Hood and Friar Tuck, which had a number of nasty payloads, such as jamming the card reader with "lace cards". The other problem was that each process watched for the presence of the other, so that on terminating a process the other would immediatly restart it. The monitor being patched meant that these processes also restarted after a reboot.
Suffice to say that as soon as the manufacturer was hit themselves by a security exploit, the hole got plugged really quick.
I have also had experience of this mechanism myself. In the past I used to communicate with the manufacturers of a defunct email system using their own product. It was possible to create a message with particular properties which would confuse the server's mail process, causing it to crash. On restart the process would immediatly crash. Sending a faulty message to the manufacturer got them to fix it PDQ!
My main ISP email address (not the hotmail one quoted here) appeared on a far-eastern mail list when I first signed up with them, 3 years or so ago.
I get around a dozen emails a week, most of which are in Chinese and therefore unreadable.
However I do get some genunine emails from the.tw and.hk domains which I'd like to keep, so a simple filter rule won't work for me. I also have no idea at present if my ISP can implement mail filtering rules.
Regarding HTML mail, I subscribe to a manually collated specialist mailing list which sends me an HTML mail message daily.
Due to the nature of the content, plain text won't really work, as it can contain tables. Unfortunatly last time I looked at the source of such a message, it was generated with a variant of Outlook, so it contains some MS specific tags.
Is there any decent Linux mail client that can handle HTML mail properly? Netscape Communicator fails to handle some of the tags at all.
I sounds like the stupid idea of someone who used to work for the same company as myself. A typical clueless suit.
He was supposedly leading our solutions's provider into new directions, but had zero technical skills. The company wasted 1000s of pounds sterling promoting ideas like this. One was signature recognition software which just didn't work. When I mentioned digital signatures he just looked at me blankly!
When I worked in network support, I had a lot of customers who relied upon old, unmaintained, software. Some of this software was over 10 years old, yet usually worked with no problems. Things like network based bulletin boards, menuing systems, 3270 print gateways and so on. It was used because no replacement could be found.
Then the suits got concerned over Year 2000 compliancy. Despite the fact that this software did no date related work, since no Year 2000 compliancy statement could be found, the software had to be junked.
There was the case with some of this software that access to the source code could have extended its life further. Some of the software was linked against a DOS networking library, which mutated over the years. One problem was the host OS version number call, which broke when the OS went above version 6. A simple recompile against the current library was sufficient, but forcing the then current software owners to do that was impossible.
Two months ago or so I saw on usenet a Windows.EXE of dubious content masquerading as both.AVI and.MPG files.
They used the usual trick of nameing the.EXE somthing like foo.AVI.EXE, and made sure that the embedded icon colour matched that of the associated fake file type.
I dumped the file using 'strings', and it appeared to generate a fake error message regarding a missing codec, as well as a registry key to autorun a program at boot. I presume this trojan contained this code.
Electronic Boutique and Game only appear to stock one distro, and that one is Turbolinux. Others have disappeared from their shelves. Seems that they use the same wholesaler. Staff are clueless even when it comes to games software. Linux experience none existant.
Software Warehouse (or whatever they are now called) seem to carry a few distros. Also quake. Staffed full of suited prats!
The small independent PC shops seem to have vanished, or be located out of town. One I used to visit was clued up, but I believe they are no longer trading.
. The best deal I've seen, though, is in Borders bookshop in Leeds. Redhat, Mandrake, SuSe, Corel, Slackware and probably others. No Debian though, other than out of date current package with the book. Many, many books and the magazine section stocks Linux Journal, Linux Magazine and Maximum Linux. I even picked up the Perl Resource kit in a sale for £2.99.
I've worked in some big IT shops in the UK, and I am amazed by the level of cluelessness of some of the management in these places.
IT policy is often dictated by who gives most sweateners to management, and never on technical ability.
I saw one company replace a perfectly valid office email system with another which not only wasn't anywhere as integrated with their OS, but it cost a fortune, and required addtional servers and administration staff. The decision to go for this product was entirely due to the MD of this company sharing a flight back from abroad with the head of sales of this software house.
I've also seen networks designed by management, and I had to remotely support a good number of them. Non-existant security, uneccesary traffic over saturated links, bad choice of hardware, the list goes on.
One person I worked for even dictated that company policy forbode the use of Pentium Processors, dictating that 486 were the company standard for everything. He also preferred Arcnet over Ethernet, despite the fact that it ran at a quarter of the speed and had other serious flaws. He later relented on the Pentium dictate, but tried to get us to use the cheapest PCs possible as file servers. I had to smuggle a high-end Compaq *INTO* the machine room when it came to replace the server.
UK management are *STUPID*. Why else has NT been adopted almost everywhere. Linux unfortunatly hasn't got a chance. And don't get me onto the subject of recruitment policy!
A few years ago, the IRA used to have fun blowing up parts of the City of London, trying to cause as much chaos as possible.
One of my collegues had to attend the clean-up operation of one such attack at a large financial institution, roughly 7-8 years ago. This company used Banyan CNS servers, which were huge 386 or 486 ISA machines with integrated UPS. The case consisted of an aluminimum frame with removable side panels and top. Due to the integrated UPS, they had to be strong.
This site had suffered severe structural damage during the attack, so much so that at least 2 of these servers were acting as supports for the rest of the building. What's more, they still worked after the event, although a bit dusty.
Banyan (remember them?) used to do the same with their VINES NOS. Based on AT&T's PC UNIX, root access was only available to one of Banyan's engineers, and was only needed when things went badly wrong, such as when the filesystem filled or became corrupted. A normal server administrator would never need root.
They later relented and gave root to customers. This was helpful, as at the same time the OS became a bit more tricky to maintain, mainly due to the removal of the need for their dongles.
BTW I knew of at least 2 mechanisms for getting root on any standard Banyan server, plus most of the root passwords, and was able to poke around a few years before root access was made public. I once spent several hours sat at the console of an old server runnign fsdb to fix a really mangled filesystem.
Performance is going to be terrible.
on
AtheOS
·
· Score: 1
First I noticed that he was using an NE2000. In this day and age should be avoided - these cards use PIO which eats into CPU utilisation. These cards were designed when 286s running single tasking DOS network stacks were the norm. They should have died out years ago.
Then I saw that it uses BIOS calls for disk access- ugh!
I've always wanted to get into OS hacking in a big way, so as soon as I get some spare disk capacity I'll consider helping them out.
They were always bound to fail - from personal experience this kind of support organisation can't no longer survive.
I spent many years working for a company offering specialist technical support on a number of platforms to many UK companies.
This was before the internet was popular, when the only resources we had were the manufacturer's technical bulletins, patch release notes, and our own experience. We flourished, because no-one else offered such a proficient service. We must have had 60+ customers, logging fairly technical support calls.
We'd have some very obscure problems to resolve, but always acheived a resolution.
Later we'd get a few other resources to work from, such as messages posted on Compuserve (remember them!).
Then in 1995, customers started to get internet access. Gradually customers use us less and less, as they were able to post to the relevant newsgroup or mailing list, or search dejanews.
No-one would use us any more to obtain patches or fixes. A savvy sysadmin could avoid using us at all, except when things went really wrong and Guru type calls were needed. For the last year there, most of my calls were of this nature.
By the beginning of 1998 we only were still going as we had gained a number of customers from our rival, as they had already decided it was no longer paying. By the end of the year we had one major customer left, and a couple of smaller sites. The company moved into facilities management and bespoke software to stay afloat.
Last year I spent four months working with a project that used Notes/Domino (why can't Lotus make up their mind) as the web server, and based upon Notes 4.6.
Based on mid-range hardware, the software ground the system to its knees, even if only a handful of users were using it simultaneously, as often happend during in-house testing.
Given that the whole idea behind this product was to act as a departmental Web solution without using any other Notes features, choosing Notes/Domino was a bad idea. It stemmed down to some 3rd party software that was only available as Windows.DLLs, and the head of the organisation believing that Notes could do everything.
Since leaving them, I've tried to avoid Notes.
Oh - and of the Web Sites, one was a low usage council infomation site. It took me many attempts to connect over a number of days, and I finally got through in the early hours of the morning. The rest of the time I connected I got something, but throughput was too slow to be usable. Other hosts on the same subnet did respond quickly, though.
So the develop this hardware to run Domino, whis is the most CPU intensive memory intensive pathetic excuse I've seen for a webserver. Plus you need to know a lot about Lotus Notes to get the thing to work at all.
Why is it that on a couple of occasions when a Web site I've tried to get to has been very slow to respond, further investigations reveal it is running Domino.
Disabling all files with a specific extension is a bit luserish, and is nearly as bad as some of the ad-hoc fixes implemented by clueless IT departments when ILV appeared. To implement this, they need some means of veifying the integrity of a file.
Only last week I received an executable via Email that I wanted to run - it contained updated hardware device drivers. If this fix had been in place, I wouldn't have been able to use it.
The various Windows platforms support embedded digital signatures in executable files. This driver update was one such one that was signed, as is becoming common practice these days.
If MS$ really knew what they were doing, then they could harness this technique to only allow execution of trusted binaries.
For example, the automated Windows update feature of IE 5 checks the validity of a program's signature before running the installation process.
So it has taken almost 2 weeks for BillyBoy & Co. to announce that a fix is in production.
How much longer will it take before the final working fix is released? Could be months, going on previous form.
Then again, it couldn't be as bad as Borderware's firewall. They once had a serious security problem, and annouced a fix would be available in 6 months(!).
I hate having to move either hand off the keyboard to operate the mouse, and have considered creating my own pointing device, capable of being used without moving the fingers.
I first considered a foot operated mouse, but discovered that in a sendentary position, lateral movement could be a problem.
I then hit on the idea of a mouth based pointing device. Used the lips or tongue to move the pointer around, and use sucking or blowing instead of clicking. Full screen movement, and both hands still on the keyboard.
Slashdot Mirror
Check the /usr/src/linux/MAINTAINERS for a full list.
In the case of my sound card driver, there was no maintainer. As the driver at the time supported at least 4 different chipsets, and I was adding support for another, I had no means to verify that my changes were indeed correct, as I was unable to locate any of the other models of that card. I did have the specs for the other chipsets, but I've been burnt by incorrect specs in the past. Luckily so far no-one has reported that my changes have broken things. But 2.4.0 hasn't gone live yet!
Maintainers are useful when multiple changes are being made to the same code, and the changes are incompatable with each other. There are many active maintainers within Linux, but still vast ranges of code without anyone to oversee it.
Of football supporters discussing Arsenal.
We have counties named Essex, and Sussex and historically though still used Middlesex and Wessex. Woops!
Award this sysadmin the BOFH award - Baldrick Operator From Hell (I have a cunning plan!).
I worked somewhere where a commercial filter the clowns into our system stopped access to well-known pron sites such as www.novell.com.
Usually this is due to the original author no longer having access to working hardware, or a platform that the hardware will work in.
Last year I tried to investigate why my sound card's Midi port wouldn't work with the drivers in the kernel. My sound card was on a weird daughterboard and not easy to remove without a hacksaw. Once I finally pulled the thing out, I dicovered that the main chips model number was later than any of those listed in the driver.
An email to the Linux Kernel mailing list received one reply, from Alan Cox, stating that no-one was now maintaining that code. In the end I amended the code myself after reverse engineering a DOS device driver. It didn't help that no specs for this particular chip were available.
Simillary the UMSDOS filesystem was broken in the 2.1.x kernel code for over 50 revisions, as the addition of dentries broke it big style. Once someone else took over the code things started to work, but it wasn't really reliable until very recently.
This relates to a system around 25 years ago, which had a severe security problem, which the vendors refused to fix. The security hole affected the monitor process, and made it possible to patch into it, IIRC.
Some hackers, disgruntled over the laxness of security, decided to exploit the problem on one of the manufacturers own systems. This installed two processes, called Robin Hood and Friar Tuck, which had a number of nasty payloads, such as jamming the card reader with "lace cards". The other problem was that each process watched for the presence of the other, so that on terminating a process the other would immediatly restart it. The monitor being patched meant that these processes also restarted after a reboot.
Suffice to say that as soon as the manufacturer was hit themselves by a security exploit, the hole got plugged really quick.
I have also had experience of this mechanism myself. In the past I used to communicate with the manufacturers of a defunct email system using their own product. It was possible to create a message with particular properties which would confuse the server's mail process, causing it to crash. On restart the process would immediatly crash. Sending a faulty message to the manufacturer got them to fix it PDQ!
Simple test - how long does it take to compute a Seti@Home work-unit?
Intel FUD, perhaps?
I get around a dozen emails a week, most of which are in Chinese and therefore unreadable.
However I do get some genunine emails from the .tw and .hk domains which I'd like to keep, so a simple filter rule won't work for me. I also have no idea at present if my ISP can implement mail filtering rules.
Now, what is the optimum algorithm for calculatin Beck's Coefficient during Knip?
Due to the nature of the content, plain text won't really work, as it can contain tables. Unfortunatly last time I looked at the source of such a message, it was generated with a variant of Outlook, so it contains some MS specific tags.
Is there any decent Linux mail client that can handle HTML mail properly? Netscape Communicator fails to handle some of the tags at all.
He was supposedly leading our solutions's provider into new directions, but had zero technical skills. The company wasted 1000s of pounds sterling promoting ideas like this. One was signature recognition software which just didn't work. When I mentioned digital signatures he just looked at me blankly!
Then the suits got concerned over Year 2000 compliancy. Despite the fact that this software did no date related work, since no Year 2000 compliancy statement could be found, the software had to be junked.
There was the case with some of this software that access to the source code could have extended its life further. Some of the software was linked against a DOS networking library, which mutated over the years. One problem was the host OS version number call, which broke when the OS went above version 6. A simple recompile against the current library was sufficient, but forcing the then current software owners to do that was impossible.
They used the usual trick of nameing the .EXE somthing like foo.AVI.EXE, and made sure that the embedded icon colour matched that of the associated fake file type.
I dumped the file using 'strings', and it appeared to generate a fake error message regarding a missing codec, as well as a registry key to autorun a program at boot. I presume this trojan contained this code.
Electronic Boutique and Game only appear to stock one distro, and that one is Turbolinux. Others have disappeared from their shelves. Seems that they use the same wholesaler. Staff are clueless even when it comes to games software. Linux experience none existant.
Software Warehouse (or whatever they are now called) seem to carry a few distros. Also quake. Staffed full of suited prats!
The small independent PC shops seem to have vanished, or be located out of town. One I used to visit was clued up, but I believe they are no longer trading.
. The best deal I've seen, though, is in Borders bookshop in Leeds. Redhat, Mandrake, SuSe, Corel, Slackware and probably others. No Debian though, other than out of date current package with the book. Many, many books and the magazine section stocks Linux Journal, Linux Magazine and Maximum Linux. I even picked up the Perl Resource kit in a sale for £2.99.
IT policy is often dictated by who gives most sweateners to management, and never on technical ability.
I saw one company replace a perfectly valid office email system with another which not only wasn't anywhere as integrated with their OS, but it cost a fortune, and required addtional servers and administration staff. The decision to go for this product was entirely due to the MD of this company sharing a flight back from abroad with the head of sales of this software house.
I've also seen networks designed by management, and I had to remotely support a good number of them. Non-existant security, uneccesary traffic over saturated links, bad choice of hardware, the list goes on.
One person I worked for even dictated that company policy forbode the use of Pentium Processors, dictating that 486 were the company standard for everything. He also preferred Arcnet over Ethernet, despite the fact that it ran at a quarter of the speed and had other serious flaws. He later relented on the Pentium dictate, but tried to get us to use the cheapest PCs possible as file servers. I had to smuggle a high-end Compaq *INTO* the machine room when it came to replace the server.
UK management are *STUPID*. Why else has NT been adopted almost everywhere. Linux unfortunatly hasn't got a chance. And don't get me onto the subject of recruitment policy!
One of my collegues had to attend the clean-up operation of one such attack at a large financial institution, roughly 7-8 years ago. This company used Banyan CNS servers, which were huge 386 or 486 ISA machines with integrated UPS. The case consisted of an aluminimum frame with removable side panels and top. Due to the integrated UPS, they had to be strong.
This site had suffered severe structural damage during the attack, so much so that at least 2 of these servers were acting as supports for the rest of the building. What's more, they still worked after the event, although a bit dusty.
Can anyone beat that..
They later relented and gave root to customers. This was helpful, as at the same time the OS became a bit more tricky to maintain, mainly due to the removal of the need for their dongles.
BTW I knew of at least 2 mechanisms for getting root on any standard Banyan server, plus most of the root passwords, and was able to poke around a few years before root access was made public. I once spent several hours sat at the console of an old server runnign fsdb to fix a really mangled filesystem.
Then I saw that it uses BIOS calls for disk access- ugh!
I've always wanted to get into OS hacking in a big way, so as soon as I get some spare disk capacity I'll consider helping them out.
Birmingham City Council.
I spent many years working for a company offering specialist technical support on a number of platforms to many UK companies.
This was before the internet was popular, when the only resources we had were the manufacturer's technical bulletins, patch release notes, and our own experience. We flourished, because no-one else offered such a proficient service. We must have had 60+ customers, logging fairly technical support calls.
We'd have some very obscure problems to resolve, but always acheived a resolution.
Later we'd get a few other resources to work from, such as messages posted on Compuserve (remember them!).
Then in 1995, customers started to get internet access. Gradually customers use us less and less, as they were able to post to the relevant newsgroup or mailing list, or search dejanews.
No-one would use us any more to obtain patches or fixes. A savvy sysadmin could avoid using us at all, except when things went really wrong and Guru type calls were needed. For the last year there, most of my calls were of this nature.
By the beginning of 1998 we only were still going as we had gained a number of customers from our rival, as they had already decided it was no longer paying. By the end of the year we had one major customer left, and a couple of smaller sites. The company moved into facilities management and bespoke software to stay afloat.
Based on mid-range hardware, the software ground the system to its knees, even if only a handful of users were using it simultaneously, as often happend during in-house testing.
Given that the whole idea behind this product was to act as a departmental Web solution without using any other Notes features, choosing Notes/Domino was a bad idea. It stemmed down to some 3rd party software that was only available as Windows .DLLs, and the head of the organisation believing that Notes could do everything.
Since leaving them, I've tried to avoid Notes.
Oh - and of the Web Sites, one was a low usage council infomation site. It took me many attempts to connect over a number of days, and I finally got through in the early hours of the morning. The rest of the time I connected I got something, but throughput was too slow to be usable. Other hosts on the same subnet did respond quickly, though.
So the develop this hardware to run Domino, whis is the most CPU intensive memory intensive pathetic excuse I've seen for a webserver. Plus you need to know a lot about Lotus Notes to get the thing to work at all.
Why is it that on a couple of occasions when a Web site I've tried to get to has been very slow to respond, further investigations reveal it is running Domino.
Only last week I received an executable via Email that I wanted to run - it contained updated hardware device drivers. If this fix had been in place, I wouldn't have been able to use it.
The various Windows platforms support embedded digital signatures in executable files. This driver update was one such one that was signed, as is becoming common practice these days.
If MS$ really knew what they were doing, then they could harness this technique to only allow execution of trusted binaries.
For example, the automated Windows update feature of IE 5 checks the validity of a program's signature before running the installation process.
BTW, is the self extracting Kerberos spec signed?
How much longer will it take before the final working fix is released? Could be months, going on previous form.
Then again, it couldn't be as bad as Borderware's firewall. They once had a serious security problem, and annouced a fix would be available in 6 months(!).
I first considered a foot operated mouse, but discovered that in a sendentary position, lateral movement could be a problem.
I then hit on the idea of a mouth based pointing device. Used the lips or tongue to move the pointer around, and use sucking or blowing instead of clicking. Full screen movement, and both hands still on the keyboard.
My other solution was to grow a third arm..