to/etc/exim.conf in the localuser and userforward directors, then user-anything@domain will be delivered to user. In the user's.forward file, $local_part_suffix will match the section after the hyphen which lets you do stuff like
if $local_part_suffix is "bar" then save/dev/null endif
If you want, you can set up a new director that requires a suffix and then treat suffixes differently from unsuffixed mail (pass it to a special forward file, that sort of thing)
The main problem is attempting to provide scalability to large numbers of processors while not compromising performance at the low end. Solaris has very fine-grained locking, which allows it to scale up to 100+CPUs (though to the best of my knowledge, Sun don't actually have any 15Ks with that many CPUs in themselves - they're having to use ours to track down the kernel bugs we keep hitting...)
Now, that locking takes kernel time. Kernel time is time that's taken away from applications. As a result, while you win at the high end you lose out somewhat at the low end, and right now more than 95% of the market is at the low end (4 processors or fewer). Attempting to make a Unix that works well on the desktop and on the massive enterprise server or supercomputer is a really, really hard problem.
I regard NFS as venerable but deprecated. Access rights based on uids?
NFS 4 fixes that.
Besides, I want network filesystems to be distributed and replicating.
Oddly, though, most people don't. Seriously, I've worked with quite large NFS setups (hundreds of machines) - people wanted bigger quotas, not a distributed filesystem.
NFS version 4 is the first with Kerberos authentication in the spec, rather than it being a Sun extension. Linux has an NFS 4 implementation in 2.5 - it ought to start appearing elsewhere before too long.
The performance of *NFS* blows chunks, full stop. Even with NFS from a Sun client to a Sun server, you'll still be significantly better off using ftp than cp as far as speed goes. Tweaking your NFS mount options makes a big difference - make sure that rsize and wsize are bumped to at least 8192. Having attribute caching on will kill performance, too.
This is inevitable with NFS - there's no provision for determining whether the server is down or whether it's just taking time to respond. The same will occur whichever OS you're using as the server. Don't use soft mounts, as they'll cause unpredictable breakage - mounting the filessystems with the intr option instead allows you to kill individual processes that are blocked on the "dead" mount.
Yes. XRender only supports one layer of trasparency - if you have a transparent XRendered object (such as a cursor), it'll show the object underneath. However, if the object underneath is also transparent, you won't be able to see through both layers to the third layer underneath. As a result of this, transparent windows don't work too well yet (though probably still better than the traditional hack of grabbing the X backdrop, shading it and pasting it in) - but since nobody is really using transparent windows, a transparent cursor is unlikely to highlight this issue.
But the people they represent (lets face it, usability experts represent the average computer user. Experienced unix people either learn what they're given or make it like what they know) aren't the people that use linux!
And the average computer user is never going to use Linux unless there's a desktop that they find easy to use.
Most switches can also be configured to accept connections only from a certain MAC address
But that's no use in this case, because you want unregistered MAC addresses to be able to communicate with the registration server.
Most switches allow you to lock out ports when something like this happens
That's more interesting. There are switches that have sufficient intelligence to allow any address from a certain range to be used but only an address that is tied to a specific MAC address from another range?
What mechanism prevents students grabbing an unallocated address and setting it up statically, then NATting through a student machine located somewhere else on the physical network? Presumably the external gateway won't pass packets that don't match a registered MAC address?
No. You can't patent sequences of DNA. You can only patent potential uses of that DNA. So, the use of BRCA2 in a diagnostic test for predisposition to breast cancer is patented in much the same way that a test for the protein it produces being used for a diagnostic test for predisposition to breast cancer could be patented (and probably is).
Biotechnology companies do, however, take the piss here. Upon finding a gene and gaining some idea about its function, they have a tendancy to file several hundred patents covering every possible uesful application of that gene.
Rosalind Franklin performed some important work that was ultimately built upon by Crick, Watson and Wilkins. Given more time she'd probably have reached the same conclusions, but the others got there first.
In science, the people who make the final discovery get more credit than the people who did the work that made this discovery possible. Chauvinism has nothing to do with it.
(On the other hand, Watson is one of the less pleasant people that I've had the poor fortune to meet)
Intel want ACPI to be adopted by as many OSes as possible, and they want it to work. Releasing their code under a license that allows OS vendors to simply integrate it into their own OS increases compatibility (the Windows ACPI interpreter has a somewhat different interpretation of the standards to the Intel one) and is plainly in their own business interests.
There are situations where a BSD-style license is preferable to the GPL. This is one of them.
Nobody forces you to use the roads, but laws exist to make it less likely that you'll be killed while doing so. When something is considered to be useful to society, we enact laws to keep it useful when it's being threatened by something else. Email is considered to be useful, and spam threatens it in a way that isn't happening in other fields. Telephones aren't rendered significantly less useful or more expensive by advertising calls. The mail system isn't rendered significantly less useful or more expensive by junk mail. Pubs aren't rendered significantly less useful or more expensive by people talking to you in them.
POP before send is a hack to get around the poor level of authenticated SMTP support in most clients. A correctly configured SMTP sever will only relay for clients with IP addresses in the local network - authenticated SMTP or POP before send allow people who aren't on the local network to relay mail through the SMTP server. This has very little to do with spam - POP before send just allows you to do something that wouldn't otherwise be possible without running an open relay. How on earth would it prevent someone from forging somebody else's email address? There's no way to pass that authentication information to remote machines, and POP before send generally allows you to use arbitrary email addresses once you've authenticated.
Well, you'll be bouncing a huge amount of legitimate mail. Outgoing mail servers are not necessarily listed in the MX records for the domain in question - nor does the reverse DNS necessarily match the domain. I send mail from home with an email address whose domain's MX record points at a machine some distance away - your description of your filtering would mean that my mail to you would be bounced.
ok. one example... it costs relatively nothing (maybe some electricity and a few big speakers) to send oral communication over the entire Central Park. on the other hand, the recipiants spend their time listening to the message. time is of value. if you didn't like the message or want to hear it, then you've wasted your time (in your opinion). if you did like the message then the time wasn't wasted (email from spammer.vs. email from a friend).
You can simply ignore it at the point where you realise that it's of no interest. If large numbers of people were all shouting at you in a manner identical to the way your friends communicated with you (and doing the same to a large number of other people simultaneously), you might expect something to be done about it.
it could also easily be argued that in the case of spam email, the recipiants costs are extremely questionable. hardware resources? ISP bill? network administration? these are all normal things you have any way. spam is just something you also get because you contine to log into an email server and check messages. maybe use a communication medium that isn't so flawed to allow such abuse.
Please don't attempt to argue that the time spend by a sysadmin dealing with tidying up after a large spam run forges his domain is of questionable value. Spam imposes a large load on network administrators that wouldn't otherwise exist. Spam costs me large numbers of CPU cycles that could be spent on something else (SpamAssassin has filtered out 90MB of spam in the past 8 months for me). Spam is responsible for companies buying new, faster hardware in order to reduce the amount of spam their users have to deal with, and responsible for them paying people to set them up and run them. If the popularity of spam increases by any significant level, the usefulness of email as a business or general comunications tool will be greatly reduced. In pretty much every single other example of unsolicited communication, you don't have this.
They don't - they'll use a forged envelope sender address so they don't receive the bounces. You don't have to modify SpamAssassin to bounce things it thinks are spam, though - just change your filtering to bounce tagged mails. Remember that in the process you'll probably be increasing the number of bounces that some poor innocent has in their mailbox.
If you don't allow people to change the email address that their client uses, how do you propose to let people actually configure the email client to use their own address?
In many cases, your first suggestion is carried out (with the result that spammers use real domains rather than made up ones). The second would be fairly impractical - my mail server's IP address resolves to a hostname that has nothing to do with the domain used for my email. People may send email from home through their ISP's mail server using their work address. Vanity domains would become significantly less useful.
A lot of spam is sent via open proxies on broadband connections in order to help the spammers make it harder to track them down. The usual response to this (if there's a response at all) is to cut the user off.
spam spam spam. if spam should be illegal, so should any form of unsolicited communication. that includes conversing to persons without their permission at the local pub.
Spam is grossly different to most other forms of unsolicited communication in one simple respect - the total cost to the recipiants is hugely larger than the total cost to the sender. This isn't true of (say) unsolicited email from an individual directly to you, unsolicted junk mail, unsolicited telephone calls or unsolicited personal conversation.
In general, it's not a good idea to accept mail unless you think you can correctly generate a bounce message if you fail to deliver it. As a result, many mail servers will refuse to accept mail if the
MAIL FROM:
section of the SMTP exchange doesn't include a domain that exists. Some will go further and do some checks to see if the localpart exists, too. If the spammers want to get to as many addresses as possible, they have to use a real address rather than a made up one. In some cases, they'll pick the address of someone who's irritated them (anti-spammers, for instance).
Slashdot Mirror
Add
/etc/exim.conf in the localuser and userforward directors, then user-anything@domain will be delivered to user. In the user's .forward file, $local_part_suffix will match the section after the hyphen which lets you do stuff like
/dev/null
suffix = -*
suffix_optional
to
if $local_part_suffix is "bar"
then
save
endif
If you want, you can set up a new director that requires a suffix and then treat suffixes differently from unsuffixed mail (pass it to a special forward file, that sort of thing)
The main problem is attempting to provide scalability to large numbers of processors while not compromising performance at the low end. Solaris has very fine-grained locking, which allows it to scale up to 100+CPUs (though to the best of my knowledge, Sun don't actually have any 15Ks with that many CPUs in themselves - they're having to use ours to track down the kernel bugs we keep hitting...)
Now, that locking takes kernel time. Kernel time is time that's taken away from applications. As a result, while you win at the high end you lose out somewhat at the low end, and right now more than 95% of the market is at the low end (4 processors or fewer). Attempting to make a Unix that works well on the desktop and on the massive enterprise server or supercomputer is a really, really hard problem.
What version? If you try mounting an NFS export without the server running portmap, it's not going to get you terribly far even with a Linux client...
I regard NFS as venerable but deprecated. Access rights based on uids?
NFS 4 fixes that.
Besides, I want network filesystems to be distributed and replicating.
Oddly, though, most people don't. Seriously, I've worked with quite large NFS setups (hundreds of machines) - people wanted bigger quotas, not a distributed filesystem.
NFS version 4 is the first with Kerberos authentication in the spec, rather than it being a Sun extension. Linux has an NFS 4 implementation in 2.5 - it ought to start appearing elsewhere before too long.
The performance of *NFS* blows chunks, full stop. Even with NFS from a Sun client to a Sun server, you'll still be significantly better off using ftp than cp as far as speed goes. Tweaking your NFS mount options makes a big difference - make sure that rsize and wsize are bumped to at least 8192. Having attribute caching on will kill performance, too.
This is inevitable with NFS - there's no provision for determining whether the server is down or whether it's just taking time to respond. The same will occur whichever OS you're using as the server. Don't use soft mounts, as they'll cause unpredictable breakage - mounting the filessystems with the intr option instead allows you to kill individual processes that are blocked on the "dead" mount.
Alpha blended cursors, but not true transparency?
Yes. XRender only supports one layer of trasparency - if you have a transparent XRendered object (such as a cursor), it'll show the object underneath. However, if the object underneath is also transparent, you won't be able to see through both layers to the third layer underneath. As a result of this, transparent windows don't work too well yet (though probably still better than the traditional hack of grabbing the X backdrop, shading it and pasting it in) - but since nobody is really using transparent windows, a transparent cursor is unlikely to highlight this issue.
Have you actually tried to use Nautilus?
The Gnome 1 version of Nautilus was justifiably criticised for its performance, but Nautilus on Gnome 2 is usably fast on relatively low-end hardware.
But the people they represent (lets face it, usability experts represent the average computer user. Experienced unix people either learn what they're given or make it like what they know) aren't the people that use linux!
And the average computer user is never going to use Linux unless there's a desktop that they find easy to use.
Most switches can also be configured to accept connections only from a certain MAC address
But that's no use in this case, because you want unregistered MAC addresses to be able to communicate with the registration server.
Most switches allow you to lock out ports when something like this happens
That's more interesting. There are switches that have sufficient intelligence to allow any address from a certain range to be used but only an address that is tied to a specific MAC address from another range?
What mechanism prevents students grabbing an unallocated address and setting it up statically, then NATting through a student machine located somewhere else on the physical network? Presumably the external gateway won't pass packets that don't match a registered MAC address?
Because of the ability to patent squences of DNA
No. You can't patent sequences of DNA. You can only patent potential uses of that DNA. So, the use of BRCA2 in a diagnostic test for predisposition to breast cancer is patented in much the same way that a test for the protein it produces being used for a diagnostic test for predisposition to breast cancer could be patented (and probably is).
Biotechnology companies do, however, take the piss here. Upon finding a gene and gaining some idea about its function, they have a tendancy to file several hundred patents covering every possible uesful application of that gene.
Rosalind Franklin performed some important work that was ultimately built upon by Crick, Watson and Wilkins. Given more time she'd probably have reached the same conclusions, but the others got there first.
In science, the people who make the final discovery get more credit than the people who did the work that made this discovery possible. Chauvinism has nothing to do with it.
(On the other hand, Watson is one of the less pleasant people that I've had the poor fortune to meet)
There are situations where a BSD-style license is preferable to the GPL. This is one of them.
Nobody forces you to use the roads, but laws exist to make it less likely that you'll be killed while doing so. When something is considered to be useful to society, we enact laws to keep it useful when it's being threatened by something else. Email is considered to be useful, and spam threatens it in a way that isn't happening in other fields. Telephones aren't rendered significantly less useful or more expensive by advertising calls. The mail system isn't rendered significantly less useful or more expensive by junk mail. Pubs aren't rendered significantly less useful or more expensive by people talking to you in them.
POP before send is a hack to get around the poor level of authenticated SMTP support in most clients. A correctly configured SMTP sever will only relay for clients with IP addresses in the local network - authenticated SMTP or POP before send allow people who aren't on the local network to relay mail through the SMTP server. This has very little to do with spam - POP before send just allows you to do something that wouldn't otherwise be possible without running an open relay. How on earth would it prevent someone from forging somebody else's email address? There's no way to pass that authentication information to remote machines, and POP before send generally allows you to use arbitrary email addresses once you've authenticated.
Well, you'll be bouncing a huge amount of legitimate mail. Outgoing mail servers are not necessarily listed in the MX records for the domain in question - nor does the reverse DNS necessarily match the domain. I send mail from home with an email address whose domain's MX record points at a machine some distance away - your description of your filtering would mean that my mail to you would be bounced.
ok. one example... it costs relatively nothing (maybe some electricity and a few big speakers) to send oral communication over the entire Central Park. on the other hand, the recipiants spend their time listening to the message. time is of value. if you didn't like the message or want to hear it, then you've wasted your time (in your opinion). if you did like the message then the time wasn't wasted (email from spammer .vs. email from a friend).
You can simply ignore it at the point where you realise that it's of no interest. If large numbers of people were all shouting at you in a manner identical to the way your friends communicated with you (and doing the same to a large number of other people simultaneously), you might expect something to be done about it.
it could also easily be argued that in the case of spam email, the recipiants costs are extremely questionable. hardware resources? ISP bill? network administration? these are all normal things you have any way. spam is just something you also get because you contine to log into an email server and check messages. maybe use a communication medium that isn't so flawed to allow such abuse.
Please don't attempt to argue that the time spend by a sysadmin dealing with tidying up after a large spam run forges his domain is of questionable value. Spam imposes a large load on network administrators that wouldn't otherwise exist. Spam costs me large numbers of CPU cycles that could be spent on something else (SpamAssassin has filtered out 90MB of spam in the past 8 months for me). Spam is responsible for companies buying new, faster hardware in order to reduce the amount of spam their users have to deal with, and responsible for them paying people to set them up and run them. If the popularity of spam increases by any significant level, the usefulness of email as a business or general comunications tool will be greatly reduced. In pretty much every single other example of unsolicited communication, you don't have this.
They don't - they'll use a forged envelope sender address so they don't receive the bounces. You don't have to modify SpamAssassin to bounce things it thinks are spam, though - just change your filtering to bounce tagged mails. Remember that in the process you'll probably be increasing the number of bounces that some poor innocent has in their mailbox.
If you don't allow people to change the email address that their client uses, how do you propose to let people actually configure the email client to use their own address?
In many cases, your first suggestion is carried out (with the result that spammers use real domains rather than made up ones). The second would be fairly impractical - my mail server's IP address resolves to a hostname that has nothing to do with the domain used for my email. People may send email from home through their ISP's mail server using their work address. Vanity domains would become significantly less useful.
A lot of spam is sent via open proxies on broadband connections in order to help the spammers make it harder to track them down. The usual response to this (if there's a response at all) is to cut the user off.
spam spam spam. if spam should be illegal, so should any form of unsolicited communication. that includes conversing to persons without their permission at the local pub.
Spam is grossly different to most other forms of unsolicited communication in one simple respect - the total cost to the recipiants is hugely larger than the total cost to the sender. This isn't true of (say) unsolicited email from an individual directly to you, unsolicted junk mail, unsolicited telephone calls or unsolicited personal conversation.
In general, it's not a good idea to accept mail unless you think you can correctly generate a bounce message if you fail to deliver it. As a result, many mail servers will refuse to accept mail if the
MAIL FROM:
section of the SMTP exchange doesn't include a domain that exists. Some will go further and do some checks to see if the localpart exists, too. If the spammers want to get to as many addresses as possible, they have to use a real address rather than a made up one. In some cases, they'll pick the address of someone who's irritated them (anti-spammers, for instance).