Have two HTB branches: one for yourself, one for good-neighbour sharing. You can set it up so the latter will be starved or almost-starved whenever you need the bandwidth. And then you can fine-tune the branches to care about TOS, etc.
Besides, traffic shaping is mandatory anyway if you want to even think about using ssh while you're downloading something.
I hate to break it to microsoft, with the glaring exception of a decent crossplatform exchange/outlook replacement, frankly I consider MS Office legacy at best.
Don't worry, it's not impossible that Sun will include a worm vector and a Fisher Price My First Mail Server(tm). They want to add enterprise-level features, it's important for leveraging paradigms to gain market-leading advantage.
Well... if you say that, I envy you the ISPs around the place where you live.
I don't even dream about IPv6 support upstream -- the place where 192.88.99.1 gets routed to is... Switzerland (and I'm in Poland). Fortunately, at least prot 41 is not blocked.
While not terribly clueful about networking myself, I'm the guy who gets called by two of four local ISPs when they want any non-trivial changes to their firewalls -- so, I at least made them provide what IPv6 could be done, and I'm making sure the firewalls do what they are supposed to do (egress, etc) but are not too strangling -- ie, they block TCP/UDP 137-139, 445, and everyone who exceeds 100 SYNs on 25 in a hour.
However, myself I'm forced to use a monopoly ISP at home. Quality of service? Since over a month, I get pings of 400-600 at the first hop between around 14.00-23.00. NO, I'M NOT SHITTING YOU. Customer service? "Everything works correctly on our side". Choice? Well, I can push the local ISPs I'm helping to connect me, it will take just two towers to get to my place.
With the upstream ISP/telco fixing things like a total outage in a timeframe of month or two, there is no way you can get an exotic service like having a protocol other than "Internet browsing" unblocked.
Wrong. A connection with a forged source address won't take any more resources than a single incoming packet, a single outgoing packet and the CPU cost of computing a cookie. That's all.
Flooding using the flooder's true address will still work, but it is trivial to block. Sure, having 100000 zombies flood a single destination will put quite a burden and will force the floodee to maintain a huge list of banned addresses, but, a single hash table on the router will alleviate anything except for bandwidth wasted. This is same as a full TCP connect() flood.
There is a TCP hack named "syn cookies", but this doesn't work very well as TCP wasn't designed to be resistant to SYN floods.
It's not that hard to find friends (usually, a friend's workmate's friend, etc, etc) who can agree to have an antenna on their roof. Usually, just giving them a sliver of bandwidth or even just a rebate in purchasing that sliver off you will make them agree.
And then you'll find yourself to be the main ISP in your town:p
I don't think of myself as of a moron -- but I do suspect that at least one of us may be a troll.
I said I already did include all the real production costs, and made a half-assed attempt at estimating the real (that is, not what the labels quote) costs of distribution. That $1.60 mean "everything till the final sound is provided, ready for pressing into CDs and/or encoding into $digital_format _plus_ all of the artists' profits".
Even if you use electronic means of distribution, the costs from now on won't be 0 -- but they certainly won't increase the price tenfold, as it's the case with RIAA.
I want my money to go into the artists' pockets, not to the middle-men. If free market is allowed, we will end up with the middle-men who can provide the most attractive marketing/packaging/delivery for the least price, instead of a greedy cartel using paralegal tricks to bash customers and any attempts at competition, or corruption-bait like the new french proposal. I'm sure that the cash paid will get "lost" in the bureaucracy.
The only part that is not complete waste is $1.60 that goes for artists' royalties. This includes recouping all of their costs, taxes, profits, etc. Everything else is just overhead.
Pressing CDs is a matter of a few cents, boxes and covers are a bit more expensive. Distribution of CDs can be way cheaper than it is the case for daily newspapers -- a CD is a bit smaller, and no one will notice if it takes weeks instead of hours to get to its destination. You can add marketing costs if you don't believe in alternate means of promotion -- just to count all the costs in the classic way. Every penny extra goes to anti-customer anti-artist parasites, the worst possible type of middle-men.
Now, the analysis above applies only if you use the old way -- CDs in plastic boxes. In comparison, using the Net reduces the distribution costs to fractions of cents per record -- and it can do all marketing for you as well.
an increased risk of exposure to faults More public review, code that tends to be of higher quality, and the ability to fix problems yourself
intellectual property claims And since when proprietary software was free from litigation?
the risk of forced disclosure of confidential code "confidential code" -- whose? If yours, you wouldn't even be able to put it there otherwise. And someone has to reread the GPL again -- no one says the gov agency in question has to distribute any source of things they use internally. If the agency in question releases some software itself -- that "confidential code" will be disclosed anyway, just in a form that is harder to read. Back in the days, I learned how to program a particular SVGA chipset by debugging through BIOS code, and my asm skills are low -- are you going to tell me that if the "confidential code" has any real value, no one will get to it anyway?
Right. That's why Oracle still has any use outside of massive databanks while there are "super" (at least "better") databases around. As it stands, free databases are not really capable of reliable operation in bigger clusters.
Speed? There is a reason why Oracle forbids publishing any benchmarks. Namely, the reason is their product simply can't stand up to MySQL speed-wise in most tasks. Indeed, there are _some_ things Oracle can do faster, and this is what gets used in official benchmarks.
Reliability? Innodb > Oracle (well, now they stiffled this). Postgres > Oracle.
Memory usage? This is downright pathethic. An untuned Oracle db takes freaking 800MB before doing anything; in another/. article an Oracle DBA claimed that I'm overestimating this and after careful tuning you can go down to 270MB. Yeah, after careful tuning. An untrimmed mysqld (most bells and whistles on) takes 18MB on start, including things brk()ed for the initial innodb checks. Postgres goes below half of this. You can claim that a modern server can afford 800MB just fine -- but what if I want to use a separate process [group] per database, or even better, a separate Xen domain? A quite basic security precaution pretty much rules out Oracle unless you want to put all your eggs into one basket. Not to mention that Oracle's approach to fixing exploits is spotty at best.
Thus: Oracle is already trumped everywhere except large databanks. This forces them to block competition in underhanded ways as opposed to actually competing.
Good to know. I'll sure check it out -- it may be a good replacement for our old test farm. As it does brutal virtualisation and emulates fake pieces of hardware, it simply can't be as fast as Xen for servers -- but since Xen doesn't do Windows, sometimes such kinds of virtualisation are needed.
So please pardon my earlier snarkiness:p It's always good to be told new things.
VMware is a fine piece of proprietary software. In fact, it is an absolute must if you have to make a non-trivial installer for a piece of a Windows software -- I can't imagine anyone reinstalling the whole damn thing every test build. And, Windows doesn't support COW. However, it's expensive. Even VMware Workstation costs as much as a new PC, and around here we can hire a person for two months for that much money. Thus, we own only a single license and gradually move away from it. Qemu+kqemu is marginally better, at the cost of no user friendliness. However, no Windows version (well, there's always Cygwin+X) and no click&droll interface means that your ordinary admin/coder/user can't make it work.
On the other hand, Xen is in a completely different league. It's not meant to be a quick&dirty tool to test an installer, Xen is pretty much supposed to be used for servers that run permanently.
As an example, the setup I'm finishing a migration to has: * nothing but firewalling in dom0 * bind9 and reverse squid in dom1 * Apache running production in dom2 * Apache running dev in dom3 * mysql in dom4 * Apache running my personal crap in dom5 Everything but dom1 is IPv6 only, too, just as an extra obfuscation layer. Script kiddies don't know IPv6, you see:p
With this kind of separation, even if you pwn what is visible to you, you are still a long way from getting to the real meat. While Xen can't manage memory dynamically well, there is next to no CPU loss (for comparison, a 2.0s-in-native test takes 2.8s in qemu+kqemu, 57s in bare qemu and 4s in vmware).
And about your problem: yeah, the documentation that goes with Xen is ABYSMAL. In fact, I couldn't get "proper" routed networking to work for the live of me. What I did, was setting up a fake bridge (_not_ tied to any actual physical interface) and tying all domUs to it. Iptables then sees xenbr0 as a regular interface, and lets you employ all your netfilter expertise in a well-known way. Yay.
I see you've been modded as "flamebait"... and I can tell who modded you down.
Russians.
Ukrainians do find squashing them with Russians offensive, but Russians, on the other hand, would do anything to take over the world. Ever heard of panslavism?
After centuries of propaganda, Russian people actually call Ukrainians, Poles and the like "ungrateful" when we fought against Russian occupation in 18-19th centuries, 1920, 1939-41, 1944-89. All because we "oppose the greatness we could achieve together with Russia".
Instead of having a black-or-white system where someone is either allowed to edit a page or completely banned from touching it, what about making it so certain low-tier users (depending on the page: anonymous, freshly created, non-admin, etc) would be unable to edit a page outright, but they would be able to submit a change which would then await a more privileged user?
Such a system would effectively remove all immutable pages, reducing the inability to edit to a mere incovenience. If you want, you could even let people view the submitted draft, having as a result a safe, reviewed version aside of unchecked changes. And utilities such as diff3 can merge simultaneous submissions in an often fully automatic way, alleviating the problem of manually combining changesets.
Thus, instead of three current levels: * free-for-all * semi-protected * immutable we could have something that can be finely tuned. A heavily vandalised page could be turned into a submission-only one, letting anyone edit the draft but requiring an admin to actually commit it; regular slightly protected pages would have anonymous/fresh users in the draft-only mode, letting anyone who had an account for more than a few days to bring in the changes.
Heck, if any change could have comments attached, it can be as good as a full expert approval system. Would that be a viable idea?
You see, Orwell's books were not fiction, but a thinly veiled image of the then-present state of Russia. The US is still far away from this, but don't worry, it's well on it's way...
Have you ever tried to use GIMP while you also had four or five OTHER programs running at the same time?
Well, you see, I try to constrain myself to only 10 desktops at a time, forcing myself to close something when I run out of them. I guess that this answers your question.
Generally, any single desktop interface breaks once you have more than 5-6 windows.
they sink right to bloody bottom of the window stack
This is a problem with Microsoft Windows(tm)(r)(C)(a)(b)(c)(e)(f)(g)(h), not with GIMP. Having separate windows lets you set them up freely; if you want to switch to some other problem, go ahead -- this is what the other desktops are for. But if you are restricted to an ancient single-desktop scheme, then yes, MDI can alleviate your problem a bit.
Thus, it's not GIMP what is a festering pile of crap. It's the lesser image editors' tendency to accomodate misdesigns of the past.
On the other hand, once you learn something, you tend to shun all different designs. I've used Turbo Pascal (WordStar keybindings) as a kid for many years, and now I use jstar exclusively, disregarding vi and emacs as crap with anal-retentive user interface, even though I do know that in theory they are superior. This can explain why people like you dislike the new, ergonomic interface and want to keep the old MDI style you're used to.
Luckily our system prevents people from fake marrying in this way by applying common sense (as judged by a court of law) to see if you were actually married. All the judge has to do is ask you to start making out with your friends to see that you're full of it.
Er, what? Show me a country where a judge can order a man to have sex with his wife.
Slashdot Mirror
Traffic shaping will do the trick just fine.
Have two HTB branches: one for yourself, one for good-neighbour sharing. You can set it up so the latter will be starved or almost-starved whenever you need the bandwidth. And then you can fine-tune the branches to care about TOS, etc.
Besides, traffic shaping is mandatory anyway if you want to even think about using ssh while you're downloading something.
I hate to break it to microsoft, with the glaring exception of a decent crossplatform exchange/outlook replacement, frankly I consider MS Office legacy at best.
Don't worry, it's not impossible that Sun will include a worm vector and a Fisher Price My First Mail Server(tm). They want to add enterprise-level features, it's important for leveraging paradigms to gain market-leading advantage.
Well... if you say that, I envy you the ISPs around the place where you live.
I don't even dream about IPv6 support upstream -- the place where 192.88.99.1 gets routed to is... Switzerland (and I'm in Poland). Fortunately, at least prot 41 is not blocked.
While not terribly clueful about networking myself, I'm the guy who gets called by two of four local ISPs when they want any non-trivial changes to their firewalls -- so, I at least made them provide what IPv6 could be done, and I'm making sure the firewalls do what they are supposed to do (egress, etc) but are not too strangling -- ie, they block TCP/UDP 137-139, 445, and everyone who exceeds 100 SYNs on 25 in a hour.
However, myself I'm forced to use a monopoly ISP at home. Quality of service? Since over a month, I get pings of 400-600 at the first hop between around 14.00-23.00. NO, I'M NOT SHITTING YOU. Customer service? "Everything works correctly on our side". Choice? Well, I can push the local ISPs I'm helping to connect me, it will take just two towers to get to my place.
With the upstream ISP/telco fixing things like a total outage in a timeframe of month or two, there is no way you can get an exotic service like having a protocol other than "Internet browsing" unblocked.
OTOH, SCTP requires only a client and a server that want to use it.
And no overzealous firewalls on the way.
Wrong. A connection with a forged source address won't take any more resources than a single incoming packet, a single outgoing packet and the CPU cost of computing a cookie. That's all.
Flooding using the flooder's true address will still work, but it is trivial to block. Sure, having 100000 zombies flood a single destination will put quite a burden and will force the floodee to maintain a huge list of banned addresses, but, a single hash table on the router will alleviate anything except for bandwidth wasted.
This is same as a full TCP connect() flood.
There is a TCP hack named "syn cookies", but this doesn't work very well as TCP wasn't designed to be resistant to SYN floods.
16 dollars minus 1 cent = $15.99
It's not that hard to find friends (usually, a friend's workmate's friend, etc, etc) who can agree to have an antenna on their roof. Usually, just giving them a sliver of bandwidth or even just a rebate in purchasing that sliver off you will make them agree.
:p
And then you'll find yourself to be the main ISP in your town
I don't think of myself as of a moron -- but I do suspect that at least one of us may be a troll.
I said I already did include all the real production costs, and made a half-assed attempt at estimating the real (that is, not what the labels quote) costs of distribution. That $1.60 mean "everything till the final sound is provided, ready for pressing into CDs and/or encoding into $digital_format _plus_ all of the artists' profits".
Even if you use electronic means of distribution, the costs from now on won't be 0 -- but they certainly won't increase the price tenfold, as it's the case with RIAA.
I want my money to go into the artists' pockets, not to the middle-men. If free market is allowed, we will end up with the middle-men who can provide the most attractive marketing/packaging/delivery for the least price, instead of a greedy cartel using paralegal tricks to bash customers and any attempts at competition, or corruption-bait like the new french proposal. I'm sure that the cash paid will get "lost" in the bureaucracy.
Whoops, here's the link to the source I used.
I think the biggest problem in the past 10 years with entertainment companies AND consumers is that each side forgets it needs the other.
Wrong.
Check out how $16-1c paid for a single record gets split (source:
$0.17 Musicians' unions
$0.80 Packaging/manufacturing
$0.82 Publishing royalties
$0.80 Retail profit
$0.90 Distribution
$1.60 Artists' royalties
$1.70 Label profit
$2.40 Marketing/promotion
$2.91 Label overhead
$3.89 Retail overhead
The only part that is not complete waste is $1.60 that goes for artists' royalties. This includes recouping all of their costs, taxes, profits, etc. Everything else is just overhead.
Pressing CDs is a matter of a few cents, boxes and covers are a bit more expensive. Distribution of CDs can be way cheaper than it is the case for daily newspapers -- a CD is a bit smaller, and no one will notice if it takes weeks instead of hours to get to its destination. You can add marketing costs if you don't believe in alternate means of promotion -- just to count all the costs in the classic way.
Every penny extra goes to anti-customer anti-artist parasites, the worst possible type of middle-men.
Now, the analysis above applies only if you use the old way -- CDs in plastic boxes. In comparison, using the Net reduces the distribution costs to fractions of cents per record -- and it can do all marketing for you as well.
So, why exactly do we need RIAA and MPAA again?
an increased risk of exposure to faults
More public review, code that tends to be of higher quality, and the ability to fix problems yourself
intellectual property claims
And since when proprietary software was free from litigation?
the risk of forced disclosure of confidential code
"confidential code" -- whose? If yours, you wouldn't even be able to put it there otherwise. And someone has to reread the GPL again -- no one says the gov agency in question has to distribute any source of things they use internally. If the agency in question releases some software itself -- that "confidential code" will be disclosed anyway, just in a form that is harder to read. Back in the days, I learned how to program a particular SVGA chipset by debugging through BIOS code, and my asm skills are low -- are you going to tell me that if the "confidential code" has any real value, no one will get to it anyway?
Not everyone needs a super-duper database.
/. article an Oracle DBA claimed that I'm overestimating this and after careful tuning you can go down to 270MB. Yeah, after careful tuning. An untrimmed mysqld (most bells and whistles on) takes 18MB on start, including things brk()ed for the initial innodb checks. Postgres goes below half of this.
Right. That's why Oracle still has any use outside of massive databanks while there are "super" (at least "better") databases around. As it stands, free databases are not really capable of reliable operation in bigger clusters.
Speed? There is a reason why Oracle forbids publishing any benchmarks. Namely, the reason is their product simply can't stand up to MySQL speed-wise in most tasks. Indeed, there are _some_ things Oracle can do faster, and this is what gets used in official benchmarks.
Reliability? Innodb > Oracle (well, now they stiffled this). Postgres > Oracle.
Memory usage? This is downright pathethic. An untuned Oracle db takes freaking 800MB before doing anything; in another
You can claim that a modern server can afford 800MB just fine -- but what if I want to use a separate process [group] per database, or even better, a separate Xen domain? A quite basic security precaution pretty much rules out Oracle unless you want to put all your eggs into one basket. Not to mention that Oracle's approach to fixing exploits is spotty at best.
Thus: Oracle is already trumped everywhere except large databanks. This forces them to block competition in underhanded ways as opposed to actually competing.
This would affect only any new versions -- and, for popular projects, you have a guarantee that a fork would spawn overnight.
Well, how exactly do you ban anything (other than un-freeing it) on a piece of Free software? :p
Right, that's a good reason to ban the Windows build of NetHack.
And how Jews answer to derogatory jokes about them?
They show the pesky goyim how to do them right.
VMWare Server is now free
:p It's always good to be told new things.
Good to know. I'll sure check it out -- it may be a good replacement for our old test farm. As it does brutal virtualisation and emulates fake pieces of hardware, it simply can't be as fast as Xen for servers -- but since Xen doesn't do Windows, sometimes such kinds of virtualisation are needed.
So please pardon my earlier snarkiness
Then it's more than twice cheaper than it used to be -- but still, 189$ more expensive than more powerful products from the competition.
VMware is a fine piece of proprietary software. In fact, it is an absolute must if you have to make a non-trivial installer for a piece of a Windows software -- I can't imagine anyone reinstalling the whole damn thing every test build. And, Windows doesn't support COW.
:p
However, it's expensive. Even VMware Workstation costs as much as a new PC, and around here we can hire a person for two months for that much money. Thus, we own only a single license and gradually move away from it.
Qemu+kqemu is marginally better, at the cost of no user friendliness. However, no Windows version (well, there's always Cygwin+X) and no click&droll interface means that your ordinary admin/coder/user can't make it work.
On the other hand, Xen is in a completely different league. It's not meant to be a quick&dirty tool to test an installer, Xen is pretty much supposed to be used for servers that run permanently.
As an example, the setup I'm finishing a migration to has:
* nothing but firewalling in dom0
* bind9 and reverse squid in dom1
* Apache running production in dom2
* Apache running dev in dom3
* mysql in dom4
* Apache running my personal crap in dom5
Everything but dom1 is IPv6 only, too, just as an extra obfuscation layer. Script kiddies don't know IPv6, you see
With this kind of separation, even if you pwn what is visible to you, you are still a long way from getting to the real meat. While Xen can't manage memory dynamically well, there is next to no CPU loss (for comparison, a 2.0s-in-native test takes 2.8s in qemu+kqemu, 57s in bare qemu and 4s in vmware).
And about your problem: yeah, the documentation that goes with Xen is ABYSMAL. In fact, I couldn't get "proper" routed networking to work for the live of me. What I did, was setting up a fake bridge (_not_ tied to any actual physical interface) and tying all domUs to it. Iptables then sees xenbr0 as a regular interface, and lets you employ all your netfilter expertise in a well-known way. Yay.
I see you've been modded as "flamebait"... and I can tell who modded you down.
Russians.
Ukrainians do find squashing them with Russians offensive, but Russians, on the other hand, would do anything to take over the world. Ever heard of panslavism?
After centuries of propaganda, Russian people actually call Ukrainians, Poles and the like "ungrateful" when we fought against Russian occupation in 18-19th centuries, 1920, 1939-41, 1944-89. All because we "oppose the greatness we could achieve together with Russia".
Instead of having a black-or-white system where someone is either allowed to edit a page or completely banned from touching it, what about making it so certain low-tier users (depending on the page: anonymous, freshly created, non-admin, etc) would be unable to edit a page outright, but they would be able to submit a change which would then await a more privileged user?
Such a system would effectively remove all immutable pages, reducing the inability to edit to a mere incovenience. If you want, you could even let people view the submitted draft, having as a result a safe, reviewed version aside of unchecked changes. And utilities such as diff3 can merge simultaneous submissions in an often fully automatic way, alleviating the problem of manually combining changesets.
Thus, instead of three current levels:
* free-for-all
* semi-protected
* immutable
we could have something that can be finely tuned. A heavily vandalised page could be turned into a submission-only one, letting anyone edit the draft but requiring an admin to actually commit it; regular slightly protected pages would have anonymous/fresh users in the draft-only mode, letting anyone who had an account for more than a few days to bring in the changes.
Heck, if any change could have comments attached, it can be as good as a full expert approval system. Would that be a viable idea?
Wait, aren't you speaking about comrade Lysenko?
You see, Orwell's books were not fiction, but a thinly veiled image of the then-present state of Russia. The US is still far away from this, but don't worry, it's well on it's way...
Have you ever tried to use GIMP while you also had four or five OTHER programs running at the same time?
Well, you see, I try to constrain myself to only 10 desktops at a time, forcing myself to close something when I run out of them. I guess that this answers your question.
Generally, any single desktop interface breaks once you have more than 5-6 windows.
they sink right to bloody bottom of the window stack
This is a problem with Microsoft Windows(tm)(r)(C)(a)(b)(c)(e)(f)(g)(h), not with GIMP. Having separate windows lets you set them up freely; if you want to switch to some other problem, go ahead -- this is what the other desktops are for. But if you are restricted to an ancient single-desktop scheme, then yes, MDI can alleviate your problem a bit.
Thus, it's not GIMP what is a festering pile of crap. It's the lesser image editors' tendency to accomodate misdesigns of the past.
On the other hand, once you learn something, you tend to shun all different designs. I've used Turbo Pascal (WordStar keybindings) as a kid for many years, and now I use jstar exclusively, disregarding vi and emacs as crap with anal-retentive user interface, even though I do know that in theory they are superior. This can explain why people like you dislike the new, ergonomic interface and want to keep the old MDI style you're used to.
Luckily our system prevents people from fake marrying in this way by applying common sense (as judged by a court of law) to see if you were actually married. All the judge has to do is ask you to start making out with your friends to see that you're full of it.
Er, what?
Show me a country where a judge can order a man to have sex with his wife.