OK so when banks go under it's mainly because of the overwhelming number of defaults. Whose fault is that?
Savers expect to be paid interest, where do you think that interest comes from? it comes from either borrowers or from the bank making investments. Both carry some risk for the bank. Would you want to save in a bank that paid no interest?
The problem is systematic risk where the chance of a whole block of loans/investments defaulting at once is higher than would neively be calculated from assuming that are independent events. In particular this happens when a large bubble bursts. Making this worse is securitisation which hides the details of what is going on from those who are actually taking on the risk.
Not only that, but apparently Windows cannot use PAE - Physical Address Extension [wikipedia.org] to address more than 4GB
Sure it can, you just have to either pay for a server edition or hack the restriction out of the kernel.
But more physical address space doesn't help here, the problem here is virtual address space for running an effective but memory hungry profile guided whole program optimisation process. Nromally 32-bit windows has a maximum of 2GB virtual memory per process (and this is one big process we are dealing with). This can be increased to 3GB at the cost of reducing the kernel address space to 1GB.
Going to a 64-bit OS (which allows 4GB of virtual address space for 32-bit processes) will buy them a little bit of time but it's not a long term soloution. Really they need a 64 to 32 cross toolchain (which according to other posts here MS do not offer) if they want to keep using profile guided optimisation as the codebase grows.
More specifically windows requires non-overlapping kernel and user address spaces. So does a regular linux kernel.
There were patches for linux ( http://lwn.net/Articles/39283/ ) that implemented a "4G/4G" system with independent user and kernel address spaces but afaict interest in them was largely lost as most newer systems became x64 capable . I doubt anything similar exists for windows.
Of course the worst type of bugs are those that turn up frequently enough to be a massive PITA but which there is no known way to reproduce. Such bugs tend to leave both users and developers MASSIVELY frustrated.
Torrents work by establishing connections either peer to peer or peer to seed. Peers request blocks of the file over those connections, download them and then check their hashes against those in the torrent file.
Peers may perform some form of tit-for-tat to punish those who download without uploading but seeds aren't trying to download anything so can't perform tit-for-tat. So it's perfectly possible to download from a torrent while refusing to upload anything.
Oh and the stats collected by the tracker are completely dependent on the honesty of the clients that report them.
IMO it's not about the total size of the game. It's about the size of the data that is needed throughout the games progress. E.g. the world map and it's associated textures and models in a game based arround missions set in a world.
To avoid huge ammounts of disc swapping either that data needs to be repeated on every disc (and hence smaller than a disc) or that data needs to be installed to a hard drive.
One option would be for MS to include a HDD in all consoles and have games with mandatory hard drive installs (and I don't mean the tiny things you see on PS3 games). Not sure how well that would go down with gamers through.
Ads that include any form of sound especially irritate me, and are the main reason i installed adblock in the first place... I also dislike garish animated ads, and those that delay access to the content but nothing is worse than sound... Especially when you can't work out which of your many browser tabs is making the noise!
And the thing is a lot of people get annoyed by obnoxious ads so they install an ad-blocker (usually ABP for firefox users) and end up blocking the non-obnoxious adds as a side affect.
The web is mostly funded by ads so IMO it's perfectly reasonable for an adblock vendor to seperate the descision of blocking obnoxious ads from the decision to block all ads.
Intel sells computer components, mostly processors, chipsets and ethernet controllers. Each computer uses one CPU, one chipset, one ethernet controller (occasionally two) etc.
The majority of those parts are sold (either directly or indirectly via a motherboard manufacturer) to OEMs who turn them into computers. Hard drives are a key component in most computers (occasionally you see a SSD only machine but it's pretty rare) so if the OEMs are supply constrained on hard drives they will reduce their purchases of everything else to match (companies HATE keeping stock these days).
Why would that be necessary, and indeed, if we are running out of IPv4 addresses, how does that help at all?
The IPv4 internet isn't going away any time soon whatever some people might want. so during the transitional period core/border routers at major ISPs (which are the only ones that need the full internet routing table now) will need to have both the IPv6 and IPv4 tables.
They could have two totally seperate sets of core/border routers for v6 and v4 but most places will probablly want to be able to run both on the same infrastructure.
I'd think that Dual Stack Lite would be the way to salvage things
DS-lite for home users and conventional tunnels for those who need real v4 IPs would save v4 addresses allow the access network to be v6 only but the core network still needs to support v4 so that the AXFR elemements and other tunnel endpoints can talk to servers on the v4 internet.
The whole idea of having such a huge space is that within an RIR, you'd have the ISPs, within the ISP, you'd have the subscribers, within the subscriber, you'd have the subnets, and within the subnets, the interface ID. Once one is within an ISP (say a/32), one can get in a customer block (a/56) and finally into a subnet (/64). It all seems to be a case of drilling down. What's it that I'm missing here?
No medium-large buisness wants to rely on a single ISP for their connectivity, to be forced to run multiple addresses in paralell for different ISPs or to be forced to change IPs when they change ISP. When one of their connections goes down they want their IPs to stop being advertised on that connection so that traffic is re-routed to their remaining connections.
The number of entries in the v6 routing table probablly will be lower than the v4 table simply because there is less legacy cruft but we are still looking at a bloody big table. Exactly how big I don't think anyone will know until deployment is complete.
I suspect this fell between the cracks of two companies.
Verizon presumablly wasn't paying for these parts (according to TFA they were replacements for suposedly failed parts under service contracts) so they probablly wouldn't have any purcahse orders or invoices for them.
Meanwhile cisco probablly didn't have information on what work each tech was doing so they could only check the reasonableness of verizons service requests as a whole, not the reasonableness of any one tech's actions.
I think prices will have to raise an order of magnitude from that level before cost of IP addresses will significantly change business behaviour.
Also all but the largest buisnesses will be able to use private addresses for internal and outbound communication and only use public addresses for machines that need to receive inbound communication.
Consider for example a router, suppose it can bring up ipv6 interfaces and even forward packets between them BUT it's doing that forwarding in software so if too large a percentage of the traffic is v6 it can't keep-up and starts dropping packets.
Now suppose that the vendor said it "supports ipv6" but all the performance promises they made related to ipv4....
I worked at a publishing company some time back, thousands of employees, each with an accessible public ip address. So we're talking gobbling up thousands of ip's when through services like NAT, etc they could have shrunk their footprint to 10 or 15 public ip's.
And then someone sends them a network abuse complaint. Unless the sender of the complaint had the forsight to record port information AND the nat saved a log of port mappings they would have no idea who the compalint was directed against.
They had grandiose ideas with IPv6 about improving routability but afaict it never really panned out that way. Afaict the idea of multihomed sites running multiple IP ranges in paralell has been pretty much abandoned and the RIRs have started handing out v6 IP blocks directly to such sites. There will be some gains as they can give an AS a bigger block at once (so less routes per AS are likely) but each routing table entry will be bigger.
Oh and of course in the transitional period (which I'd expect to be at least a decade) routers will have to handle both the IPv4 table and the IPv6 table OUCH.
Afaict hardware capabilities have grown faster than the v4 routing table so the system has kept on working. The only real problem is it's quite expensive to get a router that can import the entire table so smaller multihomed sites have to import a subset of the table and route the rest to default.
I thought there was an announcement that the IPv4 address space is now totally exhausted.
IP allocation is heirachical. The IANA assigns IPs to the RIRs, the RIRs assign IPs to ISPs and big companies, ISPs assign IPs to their customers and so-on.
Currently the IANA have run out and APNIC have run-out. The other RIRs still have IPS to hand out for now (not for much longer though).
It's true that the chances of the house losing overall on the betting is astronomically low and if it did happen they could probablly get out on a technicality (which is what I presume you are reffered to with being "legally entitled to not pay any winnings").
The thing is the casino has to pay for rent/mortgage (unless they own the property outright), property taxes, staff, energy and so on. There is a limit to how much cash they can separate from each punter so if they can't get enough punters through the door and/or those punters refuse to gamble (and hence lose) enough money the casino (and hence it's investers) will lose.
Plus remember that share prices are only loosely linked to current profitability. In theory they should be based on expected future profitability but there is a lot of emotion in there too. Just because a company makes a profit doesn't mean thier shareholders will.
Using it as a cache is one possibility, Intel bundle an implementation of that with their z68 chipset*
Doing so is potentially more efficient than deciding what to put on it yourself but it's also less predictable (what you want cached and what the computer decides to cache may not be the same) and afaict at least intel's implementation locks you into windows on the Z68 chipset. SSDs are now cheap enough that you can just use a HDD for media and a SSD for everything else without breaking the bank. Heck i'm considering not putting a HDD in my next desktop initially (especially if I HDD prices are still high when I get around to building it), I can always add one later if I need the space.
*I suspect only supporting it on z68 is an artificial restriction but I don't know for sure.
And note that while idevices and similar are "made" in china they are made by chinese contractors, not by the western companies that sell them.
Afaict the trick to dealing with china is to keep your assets (both "IP" and tangible) OUT of the country. Sure get em to fab and assemble the PCBs and put them in boxes (it's not as though they will learn anything they couldn't learn by buying your product and dissecting it) but don't put anything in that you can't afford to lose (and if you are a big company that extends to having the dealing with local companies be done in a manner that does not involve your executives traveling to china where they can be held hostage on trumped-up charges).
If you try and use SSDs for bulk storage you will waste a ton of money.
Use them for files with heavy random use patterns though (e.g. your OS and apps) and you will get a big boost in responsiveness for a relatively small outlay. Especially if the machine is a bit short on ram and can't take a ram upgrade.
That's silly. Even with the current rise in hard drive prices, SSDs are still terribly expensive by comparison.
It depends on what you are comparing.
If cost per gigabyte is your main concern then HDDs arround 2TB are still your best bet by far.
OTOH if you are comparing cost of a system drive for an office desktop things are much closer. A 60GB drive will let you install windows,office etc and still have half free. Looking at my local supplier the cheapest sata HDD* is arround £70 while a 60GB SSD is arround £80.
In the early days of the crisis there were smaller SATA drives available cheaper but at least at my local supplier they have sold out now.
*there is a 120GB IDE drive for £33 but afaict most modern motherboards don't have IDE.
No. MySQLi supports ? only, and it requires three things to be kept in sync: the statement, the list of argument types, and the list of arguments passed by reference.
ick
Porting an existing application from MySQLi to PDO MySQL to be able to use named placeholders might take a while
Is there any reason you have to port the whole app at once? afaict there is nothing stopping you porting one transaction at a time.
But it might be an easier sell if you can come up with a couple other advantages of PDO.
The only other thing I can think of is support for multiple databases but having not used mysqli I'm not a good person to ask for a comparison.
"Here are some preliminary numbers. Final numbers will probably be a bit different, but those should be a reasonable first-order approximation. Of course, your mileage may vary.
- 6GB of download. - 25GB disk space to do a single build. - 80GB disk space to build all AOSP configs at the same time. - 16GB RAM recommended, more preferred, anything less will measurably benefit from using an SSD. - 5+ hours of CPU time for a single build, 25+ minutes of wall time, as measured on my workstation (dual-E5620 i.e. 2x quad-core 2.4GHz HT, with 24GB of RAM, no SSD).
It sounds like he was building with -j16 or so (he has a hyperthreaded dual quad core so 16 virtual cores and his CPU time figure is 12 times his "wall time" figure and not every job will be using the CPU all the time) on a box with 24GB of ram. Extrapolating from there I'd think you would be more than fine with 8GB and -j4. If you have a quad core with hyperthreading then you probablly also have a machine that can take 16GB of ram
Of course if you have a dual socket monster you will want more ram so you can run at higher -j values but on single socket quad core systems you probablly won't gain much from values over 4 or 5 (a friend of mine tells me that they find stuff compiles quicker with -j5 and hyperthreading disabled than with -j9 and hyperthreading enabled. YMMV).
And unless google have been real idiots with the build system you really shouldn't need to rebuild the whole stack very often.
And even with MySQLi, you still can't use prepared statements with a variable number of ? placeholders, such as statements using operator IN or produced through a query-by-example engine [pineight.com], without a lot of pain.
Does mysqli support named placeholders? I've found writing code that generates queries isn't too bad if you have named placeholders as any desyncs between the loop that generates the query and the loop that fills in the parameters will be immediately apparent.
for example some code of mine (using PDO rather than mysqli) that generates a parameterised query.
Slashdot Mirror
OK so when banks go under it's mainly because of the overwhelming number of defaults. Whose fault is that?
Savers expect to be paid interest, where do you think that interest comes from? it comes from either borrowers or from the bank making investments. Both carry some risk for the bank. Would you want to save in a bank that paid no interest?
The problem is systematic risk where the chance of a whole block of loans/investments defaulting at once is higher than would neively be calculated from assuming that are independent events. In particular this happens when a large bubble bursts. Making this worse is securitisation which hides the details of what is going on from those who are actually taking on the risk.
Not only that, but apparently Windows cannot use PAE - Physical Address Extension [wikipedia.org] to address more than 4GB
Sure it can, you just have to either pay for a server edition or hack the restriction out of the kernel.
But more physical address space doesn't help here, the problem here is virtual address space for running an effective but memory hungry profile guided whole program optimisation process. Nromally 32-bit windows has a maximum of 2GB virtual memory per process (and this is one big process we are dealing with). This can be increased to 3GB at the cost of reducing the kernel address space to 1GB.
Going to a 64-bit OS (which allows 4GB of virtual address space for 32-bit processes) will buy them a little bit of time but it's not a long term soloution. Really they need a 64 to 32 cross toolchain (which according to other posts here MS do not offer) if they want to keep using profile guided optimisation as the codebase grows.
More specifically windows requires non-overlapping kernel and user address spaces. So does a regular linux kernel.
There were patches for linux ( http://lwn.net/Articles/39283/ ) that implemented a "4G/4G" system with independent user and kernel address spaces but afaict interest in them was largely lost as most newer systems became x64 capable . I doubt anything similar exists for windows.
At least with a 32-bit browser it crashes when it runs out of address space rather than sending the whole machine into swap-death.
Of course the worst type of bugs are those that turn up frequently enough to be a massive PITA but which there is no known way to reproduce. Such bugs tend to leave both users and developers MASSIVELY frustrated.
Torrents work by establishing connections either peer to peer or peer to seed. Peers request blocks of the file over those connections, download them and then check their hashes against those in the torrent file.
Peers may perform some form of tit-for-tat to punish those who download without uploading but seeds aren't trying to download anything so can't perform tit-for-tat. So it's perfectly possible to download from a torrent while refusing to upload anything.
Oh and the stats collected by the tracker are completely dependent on the honesty of the clients that report them.
IMO it's not about the total size of the game. It's about the size of the data that is needed throughout the games progress. E.g. the world map and it's associated textures and models in a game based arround missions set in a world.
To avoid huge ammounts of disc swapping either that data needs to be repeated on every disc (and hence smaller than a disc) or that data needs to be installed to a hard drive.
One option would be for MS to include a HDD in all consoles and have games with mandatory hard drive installs (and I don't mean the tiny things you see on PS3 games). Not sure how well that would go down with gamers through.
Ads that include any form of sound especially irritate me, and are the main reason i installed adblock in the first place... I also dislike garish animated ads, and those that delay access to the content but nothing is worse than sound... Especially when you can't work out which of your many browser tabs is making the noise!
And the thing is a lot of people get annoyed by obnoxious ads so they install an ad-blocker (usually ABP for firefox users) and end up blocking the non-obnoxious adds as a side affect.
The web is mostly funded by ads so IMO it's perfectly reasonable for an adblock vendor to seperate the descision of blocking obnoxious ads from the decision to block all ads.
Intel sells computer components, mostly processors, chipsets and ethernet controllers. Each computer uses one CPU, one chipset, one ethernet controller (occasionally two) etc.
The majority of those parts are sold (either directly or indirectly via a motherboard manufacturer) to OEMs who turn them into computers. Hard drives are a key component in most computers (occasionally you see a SSD only machine but it's pretty rare) so if the OEMs are supply constrained on hard drives they will reduce their purchases of everything else to match (companies HATE keeping stock these days).
Why would that be necessary, and indeed, if we are running out of IPv4 addresses, how does that help at all?
The IPv4 internet isn't going away any time soon whatever some people might want. so during the transitional period core/border routers at major ISPs (which are the only ones that need the full internet routing table now) will need to have both the IPv6 and IPv4 tables.
They could have two totally seperate sets of core/border routers for v6 and v4 but most places will probablly want to be able to run both on the same infrastructure.
I'd think that Dual Stack Lite would be the way to salvage things
DS-lite for home users and conventional tunnels for those who need real v4 IPs would save v4 addresses allow the access network to be v6 only but the core network still needs to support v4 so that the AXFR elemements and other tunnel endpoints can talk to servers on the v4 internet.
The whole idea of having such a huge space is that within an RIR, you'd have the ISPs, within the ISP, you'd have the subscribers, within the subscriber, you'd have the subnets, and within the subnets, the interface ID. Once one is within an ISP (say a /32), one can get in a customer block (a /56) and finally into a subnet (/64). It all seems to be a case of drilling down. What's it that I'm missing here?
No medium-large buisness wants to rely on a single ISP for their connectivity, to be forced to run multiple addresses in paralell for different ISPs or to be forced to change IPs when they change ISP. When one of their connections goes down they want their IPs to stop being advertised on that connection so that traffic is re-routed to their remaining connections.
The number of entries in the v6 routing table probablly will be lower than the v4 table simply because there is less legacy cruft but we are still looking at a bloody big table. Exactly how big I don't think anyone will know until deployment is complete.
I suspect this fell between the cracks of two companies.
Verizon presumablly wasn't paying for these parts (according to TFA they were replacements for suposedly failed parts under service contracts) so they probablly wouldn't have any purcahse orders or invoices for them.
Meanwhile cisco probablly didn't have information on what work each tech was doing so they could only check the reasonableness of verizons service requests as a whole, not the reasonableness of any one tech's actions.
at $11.5 a pop.at $11.5 a pop.
I think prices will have to raise an order of magnitude from that level before cost of IP addresses will significantly change business behaviour.
Also all but the largest buisnesses will be able to use private addresses for internal and outbound communication and only use public addresses for machines that need to receive inbound communication.
The problem is the definition of "supports".
Consider for example a router, suppose it can bring up ipv6 interfaces and even forward packets between them BUT it's doing that forwarding in software so if too large a percentage of the traffic is v6 it can't keep-up and starts dropping packets.
Now suppose that the vendor said it "supports ipv6" but all the performance promises they made related to ipv4....
I worked at a publishing company some time back, thousands of employees, each with an accessible public ip address. So we're talking gobbling up thousands of ip's when through services like NAT, etc they could have shrunk their footprint to 10 or 15 public ip's.
And then someone sends them a network abuse complaint. Unless the sender of the complaint had the forsight to record port information AND the nat saved a log of port mappings they would have no idea who the compalint was directed against.
They had grandiose ideas with IPv6 about improving routability but afaict it never really panned out that way. Afaict the idea of multihomed sites running multiple IP ranges in paralell has been pretty much abandoned and the RIRs have started handing out v6 IP blocks directly to such sites. There will be some gains as they can give an AS a bigger block at once (so less routes per AS are likely) but each routing table entry will be bigger.
Oh and of course in the transitional period (which I'd expect to be at least a decade) routers will have to handle both the IPv4 table and the IPv6 table OUCH.
Afaict hardware capabilities have grown faster than the v4 routing table so the system has kept on working. The only real problem is it's quite expensive to get a router that can import the entire table so smaller multihomed sites have to import a subset of the table and route the rest to default.
I thought there was an announcement that the IPv4 address space is now totally exhausted.
IP allocation is heirachical. The IANA assigns IPs to the RIRs, the RIRs assign IPs to ISPs and big companies, ISPs assign IPs to their customers and so-on.
Currently the IANA have run out and APNIC have run-out. The other RIRs still have IPS to hand out for now (not for much longer though).
It's true that the chances of the house losing overall on the betting is astronomically low and if it did happen they could probablly get out on a technicality (which is what I presume you are reffered to with being "legally entitled to not pay any winnings").
The thing is the casino has to pay for rent/mortgage (unless they own the property outright), property taxes, staff, energy and so on. There is a limit to how much cash they can separate from each punter so if they can't get enough punters through the door and/or those punters refuse to gamble (and hence lose) enough money the casino (and hence it's investers) will lose.
Plus remember that share prices are only loosely linked to current profitability. In theory they should be based on expected future profitability but there is a lot of emotion in there too. Just because a company makes a profit doesn't mean thier shareholders will.
Using it as a cache is one possibility, Intel bundle an implementation of that with their z68 chipset*
Doing so is potentially more efficient than deciding what to put on it yourself but it's also less predictable (what you want cached and what the computer decides to cache may not be the same) and afaict at least intel's implementation locks you into windows on the Z68 chipset. SSDs are now cheap enough that you can just use a HDD for media and a SSD for everything else without breaking the bank. Heck i'm considering not putting a HDD in my next desktop initially (especially if I HDD prices are still high when I get around to building it), I can always add one later if I need the space.
*I suspect only supporting it on z68 is an artificial restriction but I don't know for sure.
And note that while idevices and similar are "made" in china they are made by chinese contractors, not by the western companies that sell them.
Afaict the trick to dealing with china is to keep your assets (both "IP" and tangible) OUT of the country. Sure get em to fab and assemble the PCBs and put them in boxes (it's not as though they will learn anything they couldn't learn by buying your product and dissecting it) but don't put anything in that you can't afford to lose (and if you are a big company that extends to having the dealing with local companies be done in a manner that does not involve your executives traveling to china where they can be held hostage on trumped-up charges).
If you try and use SSDs for bulk storage you will waste a ton of money.
Use them for files with heavy random use patterns though (e.g. your OS and apps) and you will get a big boost in responsiveness for a relatively small outlay. Especially if the machine is a bit short on ram and can't take a ram upgrade.
That's silly. Even with the current rise in hard drive prices, SSDs are still terribly expensive by comparison.
It depends on what you are comparing.
If cost per gigabyte is your main concern then HDDs arround 2TB are still your best bet by far.
OTOH if you are comparing cost of a system drive for an office desktop things are much closer. A 60GB drive will let you install windows,office etc and still have half free. Looking at my local supplier the cheapest sata HDD* is arround £70 while a 60GB SSD is arround £80.
In the early days of the crisis there were smaller SATA drives available cheaper but at least at my local supplier they have sold out now.
*there is a 120GB IDE drive for £33 but afaict most modern motherboards don't have IDE.
No. MySQLi supports ? only, and it requires three things to be kept in sync: the statement, the list of argument types, and the list of arguments passed by reference.
ick
Porting an existing application from MySQLi to PDO MySQL to be able to use named placeholders might take a while
Is there any reason you have to port the whole app at once? afaict there is nothing stopping you porting one transaction at a time.
But it might be an easier sell if you can come up with a couple other advantages of PDO.
The only other thing I can think of is support for multiple databases but having not used mysqli I'm not a good person to ask for a comparison.
The full quote is
"Here are some preliminary numbers. Final numbers will probably be a bit different, but those should be a reasonable first-order approximation. Of course, your mileage may vary.
- 6GB of download.
- 25GB disk space to do a single build.
- 80GB disk space to build all AOSP configs at the same time.
- 16GB RAM recommended, more preferred, anything less will measurably benefit from using an SSD.
- 5+ hours of CPU time for a single build, 25+ minutes of wall time, as measured on my workstation (dual-E5620 i.e. 2x quad-core 2.4GHz HT, with 24GB of RAM, no SSD).
It sounds like he was building with -j16 or so (he has a hyperthreaded dual quad core so 16 virtual cores and his CPU time figure is 12 times his "wall time" figure and not every job will be using the CPU all the time) on a box with 24GB of ram. Extrapolating from there I'd think you would be more than fine with 8GB and -j4. If you have a quad core with hyperthreading then you probablly also have a machine that can take 16GB of ram
Of course if you have a dual socket monster you will want more ram so you can run at higher -j values but on single socket quad core systems you probablly won't gain much from values over 4 or 5 (a friend of mine tells me that they find stuff compiles quicker with -j5 and hyperthreading disabled than with -j9 and hyperthreading enabled. YMMV).
And unless google have been real idiots with the build system you really shouldn't need to rebuild the whole stack very often.
what exactly is the problem with 16GB of RAM? it's about the same as subscribing in the apple developer program for a year..
Depends if you already have the CPU/MB to support it.
But I think the 16GB figure is rather exaggerated unless you are doing frequent rebuilds of the whole stack. (see my other post for why)
And even with MySQLi, you still can't use prepared statements with a variable number of ? placeholders, such as statements using operator IN or produced through a query-by-example engine [pineight.com], without a lot of pain.
Does mysqli support named placeholders? I've found writing code that generates queries isn't too bad if you have named placeholders as any desyncs between the loop that generates the query and the loop that fills in the parameters will be immediately apparent.
for example some code of mine (using PDO rather than mysqli) that generates a parameterised query.
http://isis.philpem.me.uk/hg/isis/file/9364d6277308/app/models/searchmodel.class.php