The funniest thing about a recent trip to Vienna (in Austria) was seeing some of the local students walking around with T-Shirts with a "No kangaroo" signs upon them.
It seems that many tourists got Austria and Australia confused - quite how that can be I don't understand - and constantly asked to see the kangaroos.
Retraining is one of those things that is often used with a selective memory.
For example "You can't switch to Linux for your secretaries desktops, you'd have to retrain them!".
But then you have "Use our new Office, it has ribbons. They rock. Retraining you say? Nah, they'll pick it up in no time".
People adapt quickly. I loved Windows 2000, not too flashy, and pretty solid. But I had to relearn a lot when the company I was working for jumped to XP. Now I'd probably be stumped when dealing with a 2000 machine again.
Thankfully I've used nothing but Debian at home for many years, so no retraining or admin changes there.
Indeed. The standard "reject" or "accept" solutions are very brittle - because the any false positives are disasters as far as your users are concerned.
my solution involves rejecting mail which is determined to be SPAM at SMTP-time but also keeping a copy if it in a quarantine where it may be searched/viewed/re-delivered.
It is still brittle - but the sender knows that their message was not delivered, due to the bounce, and the recipient does have a copy in their quarantine area, if they look for it.
The way I attempted to solve the mass-mailing problem was two-fold:
The simple method: MD5 hash(text) should be unique
The complex method: Character analysis
The simple method of stopping mass-mailings was to forbid a user from sending identical messages to multiple people, just storing a hash of past messages would work well. This restriction could be relaxed if there were replies, or other heuristics. (To avoid the problem of "hello how are you today" amongst friends)
The other problem was to do letter-frequency analysis on the text of the message, and the profile of the recipient. If they were too different you could drop the message as spam. (Essentially I wanted a lightweight bayasian filter which was initially trained on the text of your profile. It worked nicely:)
As for the problem with people lying. I figured I'd sidestep that. Don't ask people their weight; they'll not be able to lie. I encourage people to upload pictures instead, as a more obvious way of seeing whether you like somebody. Sure people can hide aspects of their body they don't like, but it is a little harder, and removes the obvious lies.
The idea of stopping new messages from being read until old ones were read, replied or "resolved" via an "I'm not interested & block" button is very nice. I might steal it;)
It's a tricky problem to solve, especially as there aren't many venues where you can advertise that kind of site.
Still I wish you luck with it. It is a fascinating problem from a geek-perspective. And I love to play cupid, even if it rarely happens the thought is nice.
The idea you're suggesting; namely listing response rate has already been implemented.
I'm a Debian developer, who is interested in kink. There are three big dating sites for that kind of audience:
alt.com
bondage.com
collarme.com
These are the biggies. There are smaller ones in particular geographical locations, and focussed upon particular kinks.
I think all three suck. Alt.com & Bondage.com are commercial and hard to use unless you pay. Collarme is ful of trolls and fakes.
So, to experiment with different things I setup my own site. I put together a kink-themed website, with a geeky name, ctrl-alt-date.
Unfortunately I'm spoilt by the problem you note, and I didn't expect. Everybody goes to the big three. Sure they suck, but they are (undeniably) where the audience is. More audience == more chance of sex/hookups/relationship.
(I guess there is also something you don't mention. The audience for a dating site is very random. If you get a partner you never return - so you end up with millions of orphan accounts interfering with search results. Its a numbers game to a certain extent too - if site A has 10million members listed you go there over site B with 5 million members. Regardless of whether both have actually only got 3 million active users.)
My site is tiny <1000 users. But it does have some novel ideas coded, and more which I'd like to test if I had the numbers. For example you can simply mark your profile as unavailiable to Straight Men, and that way you never even show up on the search results for a man - perfect if you're a lesbian,for example.
It seems to me that if you're wanting to be found by a new partner you want to do two things:
Make it impossible for bad matches to find you.
Make it easy for good matches to find you.
I'm aiming more at the first point, but the second is interesting too. One idea is allowing random strangers to edit your profile, or leave suggestions on improving it in exchange for perks. THat ups the quality of the profiles at minimal cost.
I could write more about the subject, but I might be boring people - so I'll stop for now.
On a serious note - virtualisation plays a major part in computing today, if Microsoft's os is going to be virtualised it will be done on thier terms and of course deeply integrated.
Agreed.
One of the benefits of virtualised Linux is that you can run minimal VMs for different purposes. Right now I have a desktop running about 8 Xen instances of Debian, in a mixture of Stable, Testing, and Unstable.
If you imagine running Microsoft Windows for testing you'd be interested in running XP, Vista, and Server. That is most likely going to become a licensing nightmare.
Running virtual Windows should be OK itself, if your host hardware is sufficient (since Windows has GUI and is generally "busier" than Linux installs. But the licensing cost(s) will be something that don't tend to apply to the Linux virtualization.
The challenge there is to build a better bot, (but since there's chat involved in the game, you'd better get ready for a Turing test; since that isn't an option, discretion is the better part of valour).
If you were going to massive effort to write a new bot wouldn't it make sense to proxy the chat requests to an instance of Jabber, or similar?
That way you could have the bot doing bot-things, and if you get a chat a farm-operator could handle all incoming/outgoing chat queries from one central machine.
As for Episode 3 being some kind of redemption, sorry no. Granted, all the work he put into 1 and 2 reach some neat and satisfying conclusions, but he still managed to deliver a highly-flawed and (as usual) stodgy movie.
See this is where I lost most interest.
When I heard there were going to be prequels made I knew I'd have to see them. The first I thought was disappointing, but I went to see the second regardless.
After being underwhelmed with Ep1 & Ep2 I just didn't bother to see the third. When I hear people say "Weeellll it was better than the previous two" that doesn't sell it to me.
I'm disappointed I didn't see the full story-arc, but at the same time I'm more annoyed with myself that I spoiled the relatively good childhood memories with the first two.
Given the number of hijacked machines taking part in the Storm worm, for example, any popularity contest could be skewed by a maliciously motivated attacker.
The big issue with tor is that you're magnifying your exposure. By default you're vulnerable to sniffing by your ISP, and all the people they peer with till you get to your endpoint. With tor in the mix you're vulnerable to sniffing from your ISP, and any number of random people who've elected to host a tor node.
Sure you've bounced your connection around to essentially mask the source & destination from the end-point and your ISP - but you've introduce a whole load of untrusted hops as part of that.
If you care about security the idea of passing unencrypted traffic through even more random machines should scare you..
Clearly the only way we can resolve this (and simultaneously cause major problems for Best Buy) is to unpack our purchases while we stand at the counter to check the contents. I mean, obviously you can't trust Best Buy.
Funny you should say this.. At the weekend I went to buy a copy of Worms: Open Warfare 2 for the DS. At the counter I opened the box to check it was correct and discovered it had "Open Warfare 1" inside it. It would have sucked to get home (30 minute bus ride) before spotting it.
I always check games, DVDs, and CDs I buy to make sure I have the correct disks. Sometimes I find mistakes and it doesn't take long to do. (I only buy DVDs & CDs second-hand/used from small shops so there is no shrinkwrap involved.)
The thing that I don't like about this scheme, is that YCM will (wisely) withdraw all the lawsuits before focusing on making a profit out of the deal. This means that at the end of the whole 4 year mess, there will be no clear verdict stating how badly SCO has behaved.
The thing is I'm not sure they can.
Whilst they could certainly drop the lawsuit against IBM they are still going to be liable for the counter-claims IBM and Novell (and redhat?) filed against them - and they can't force those to be dropped. Especially if Novell are keen on getting the money they are owed.
Slashdot Mirror
The funniest thing about a recent trip to Vienna (in Austria) was seeing some of the local students walking around with T-Shirts with a "No kangaroo" signs upon them.
It seems that many tourists got Austria and Australia confused - quite how that can be I don't understand - and constantly asked to see the kangaroos.
Retraining is one of those things that is often used with a selective memory.
For example "You can't switch to Linux for your secretaries desktops, you'd have to retrain them!".
But then you have "Use our new Office, it has ribbons. They rock. Retraining you say? Nah, they'll pick it up in no time".
People adapt quickly. I loved Windows 2000, not too flashy, and pretty solid. But I had to relearn a lot when the company I was working for jumped to XP. Now I'd probably be stumped when dealing with a 2000 machine again.
Thankfully I've used nothing but Debian at home for many years, so no retraining or admin changes there.
Maybe an award for the number of security issues the code has historically had?
I am an off-site mail filterer, and our stats show that 99% of incoming mail is spam.
Indeed. The standard "reject" or "accept" solutions are very brittle - because the any false positives are disasters as far as your users are concerned.
my solution involves rejecting mail which is determined to be SPAM at SMTP-time but also keeping a copy if it in a quarantine where it may be searched/viewed/re-delivered.
It is still brittle - but the sender knows that their message was not delivered, due to the bounce, and the recipient does have a copy in their quarantine area, if they look for it.
Although recently a Debian Developer was critical of slicehost, and seemingly in a valid way.
Personally I host a reasonably high-traffic antispam service and I think Amazon's offering looks good, but as mentioned a little pricy.
I love the idea of adding extra nodes on-demand, but I think I'm not yet at the level where it would be a worthwhile use of my time or budget.
I'm pretty sure there are multiple strains of HIV out there, so it is entirely possible it could mutate.
Greetings from Edinburgh - remind me not to party in Glasgow again in the future! ;)
This is what I have in my ~/.emacs file for that:
(The complete file is available online.)
The way I attempted to solve the mass-mailing problem was two-fold:
The simple method of stopping mass-mailings was to forbid a user from sending identical messages to multiple people, just storing a hash of past messages would work well. This restriction could be relaxed if there were replies, or other heuristics. (To avoid the problem of "hello how are you today" amongst friends)
The other problem was to do letter-frequency analysis on the text of the message, and the profile of the recipient. If they were too different you could drop the message as spam. (Essentially I wanted a lightweight bayasian filter which was initially trained on the text of your profile. It worked nicely :)
As for the problem with people lying. I figured I'd sidestep that. Don't ask people their weight; they'll not be able to lie. I encourage people to upload pictures instead, as a more obvious way of seeing whether you like somebody. Sure people can hide aspects of their body they don't like, but it is a little harder, and removes the obvious lies.
The idea of stopping new messages from being read until old ones were read, replied or "resolved" via an "I'm not interested & block" button is very nice. I might steal it ;)
It's a tricky problem to solve, especially as there aren't many venues where you can advertise that kind of site.
Still I wish you luck with it. It is a fascinating problem from a geek-perspective. And I love to play cupid, even if it rarely happens the thought is nice.
The idea you're suggesting; namely listing response rate has already been implemented.
I'm a Debian developer, who is interested in kink. There are three big dating sites for that kind of audience:
These are the biggies. There are smaller ones in particular geographical locations, and focussed upon particular kinks.
I think all three suck. Alt.com & Bondage.com are commercial and hard to use unless you pay. Collarme is ful of trolls and fakes.
So, to experiment with different things I setup my own site. I put together a kink-themed website, with a geeky name, ctrl-alt-date.
Unfortunately I'm spoilt by the problem you note, and I didn't expect. Everybody goes to the big three. Sure they suck, but they are (undeniably) where the audience is. More audience == more chance of sex/hookups/relationship.
(I guess there is also something you don't mention. The audience for a dating site is very random. If you get a partner you never return - so you end up with millions of orphan accounts interfering with search results. Its a numbers game to a certain extent too - if site A has 10million members listed you go there over site B with 5 million members. Regardless of whether both have actually only got 3 million active users.)
My site is tiny <1000 users. But it does have some novel ideas coded, and more which I'd like to test if I had the numbers. For example you can simply mark your profile as unavailiable to Straight Men, and that way you never even show up on the search results for a man - perfect if you're a lesbian,for example.
It seems to me that if you're wanting to be found by a new partner you want to do two things:
I'm aiming more at the first point, but the second is interesting too. One idea is allowing random strangers to edit your profile, or leave suggestions on improving it in exchange for perks. THat ups the quality of the profiles at minimal cost.
I could write more about the subject, but I might be boring people - so I'll stop for now.
If it is just templates you want then no need to buy, the open source web design site has a huge number of templates.
Sure some of them have strings attached, but I've used them a lot in the past when I needed a quick-start at designing a new site.
Agreed.
One of the benefits of virtualised Linux is that you can run minimal VMs for different purposes. Right now I have a desktop running about 8 Xen instances of Debian, in a mixture of Stable, Testing, and Unstable.
If you imagine running Microsoft Windows for testing you'd be interested in running XP, Vista, and Server. That is most likely going to become a licensing nightmare.
Running virtual Windows should be OK itself, if your host hardware is sufficient (since Windows has GUI and is generally "busier" than Linux installs. But the licensing cost(s) will be something that don't tend to apply to the Linux virtualization.
If you were going to massive effort to write a new bot wouldn't it make sense to proxy the chat requests to an instance of Jabber, or similar?
That way you could have the bot doing bot-things, and if you get a chat a farm-operator could handle all incoming/outgoing chat queries from one central machine.
Are you sure there isn't a missing virtual/cabal in there somewhere?
I know Debian depends upon one ;)
Insightful? Pah!
I'm a sysadmin and I came to become one after working as a developer for a good many years.
There are the same interesting bits involved being a sysadmin, along with debugging plus you get to have a hell of a better budget!
The best part is I can still write code to automate jobs across the machines I maintain via puppet/cfengine..
Wasn't that pretty common with machines of the era?
I know that my Spectrum 48k had a nice orange ring-bound manual complete with sample programs, up to and including a table of machine code opcodes.
There was even a slim introductory manual which contained a photograph of the machines main circuit-board, with the major components labeled.
(In fact one of my biggest disappointment with the later Sinclair machines, post Amstrad buyout, was the sparseness of the manual(s)).
Here are two security reportsfrm Georgi Guninski, neither of which received the bounty offered:
See this is where I lost most interest.
When I heard there were going to be prequels made I knew I'd have to see them. The first I thought was disappointing, but I went to see the second regardless.
After being underwhelmed with Ep1 & Ep2 I just didn't bother to see the third. When I hear people say "Weeellll it was better than the previous two" that doesn't sell it to me.
I'm disappointed I didn't see the full story-arc, but at the same time I'm more annoyed with myself that I spoiled the relatively good childhood memories with the first two.
And now the punchline ... In the UK our number is 999, but nowadays 911 works too.
I remember the rationale given that many children would see it on TV and not know it wasn't supposed to apply to them...
I seem to remember that R-Type on my +2A required reloading previous levels if you died and started again at level 1. That was very annoying ..
Given the number of hijacked machines taking part in the Storm worm, for example, any popularity contest could be skewed by a maliciously motivated attacker.
The big issue with tor is that you're magnifying your exposure. By default you're vulnerable to sniffing by your ISP, and all the people they peer with till you get to your endpoint. With tor in the mix you're vulnerable to sniffing from your ISP, and any number of random people who've elected to host a tor node.
Sure you've bounced your connection around to essentially mask the source & destination from the end-point and your ISP - but you've introduce a whole load of untrusted hops as part of that.
If you care about security the idea of passing unencrypted traffic through even more random machines should scare you ..
That sounds plausible - mum, dad, and children 1-N..
Funny you should say this .. At the weekend I went to buy a copy of Worms: Open Warfare 2 for the DS. At the counter I opened the box to check it was correct and discovered it had "Open Warfare 1" inside it. It would have sucked to get home (30 minute bus ride) before spotting it.
I always check games, DVDs, and CDs I buy to make sure I have the correct disks. Sometimes I find mistakes and it doesn't take long to do. (I only buy DVDs & CDs second-hand/used from small shops so there is no shrinkwrap involved.)
The thing is I'm not sure they can.
Whilst they could certainly drop the lawsuit against IBM they are still going to be liable for the counter-claims IBM and Novell (and redhat?) filed against them - and they can't force those to be dropped. Especially if Novell are keen on getting the money they are owed.