While the fixed disk shipped by an OEM has the latest service pack installed, the CD shipped with the system was always the same as the one sold at retail - and might be a year or two out of date. My OEM told me that their agreement with MS forbid them from putting a patched OS on the CD. (Yes, they had to arrange to manufacture the CDs also, MS only supplied the hologram).
I have worked with census data, and even taken a course in data protection in DC, so I can clear up some things.
First, census has two basic products. The first shows the number of people of each race, gender income category and age category by census block.
If any category only has a few people in it, then
that category is suppressed and the people in that category are added to a nearby one. So in rural Minnesota the one hispanic family is dropped replaced with a dash and a family is added to the 'white' category. There are theoretical attacks possible on this sort of data, especially where other tables are published. There are no examples of citizens complaining that their particular information has been disclosed.
The second category of data are called 'public use files'. This is a sample of 1% of the forms, but with little or no geographic data. So in this
file you can see that one person had this income category and that race and that age etc, through all the data in the census form. But you only know the state of residence. So you couldn't locate a particular individual. And the categories are wide enough that there would be multiple people with the same information, and census has computer programs to make sure that is the case. If it isn't, they fuzz the data a bit, untill it is. So far no citizen has ever complained that their personal info has been compromised by this data.
After 10 years of dealing with vendors (SUN, IBM, SAS, NetAPP, ECCS, HP) I can only be envious of someone who expects tech support from one. I can only recall a single instance when a vendor fixed something in response to my trouble report. Even in single vendor setups, the DB support group will have no hesitation to blame the hardware group, or the OS group for problems and close the problem report as 'solved'. If there is a vendor that doesn't do that, someone please mention them here. And I don't know any vendor that would not be perfectly comfortable closing a problem report with 'That is a known bug documented in TR123456. No patch or workaround is available'. Only a few favored customers actually get tech support, the rest might as well be buying at a flee market. The best way to protect yourself is to make sure that your configuration and purpose are very similar to a favored customer, or a target market.
The advantage Capture has in creating PDFs from OCR is that when Capture is not sure of an interpretation it just puts a little picture of the word into the PDF file, instead of taking a wild guess. The resulting file is searchable using ordinary PDF tools, on any words that were successfully OCRed, but of course the little pictures aren't indexed. We have found that Capture is a bit too optimistic and thinks it has the words down when it doesn't, and this limits our use of Capture, but in principle using a combination of OCR and pictures is the way to go.
Slashdot Mirror
While the fixed disk shipped by an OEM has the latest service pack installed, the CD shipped with the system was always the same as the one sold at retail - and might be a year or two out of date. My OEM told me that their agreement with MS forbid them from putting a patched OS on the CD. (Yes, they had to arrange to manufacture the CDs also, MS only supplied the hologram).
I have worked with census data, and even taken a course in data protection in DC, so I can clear up some things. First, census has two basic products. The first shows the number of people of each race, gender income category and age category by census block. If any category only has a few people in it, then that category is suppressed and the people in that category are added to a nearby one. So in rural Minnesota the one hispanic family is dropped replaced with a dash and a family is added to the 'white' category. There are theoretical attacks possible on this sort of data, especially where other tables are published. There are no examples of citizens complaining that their particular information has been disclosed. The second category of data are called 'public use files'. This is a sample of 1% of the forms, but with little or no geographic data. So in this file you can see that one person had this income category and that race and that age etc, through all the data in the census form. But you only know the state of residence. So you couldn't locate a particular individual. And the categories are wide enough that there would be multiple people with the same information, and census has computer programs to make sure that is the case. If it isn't, they fuzz the data a bit, untill it is. So far no citizen has ever complained that their personal info has been compromised by this data.
After 10 years of dealing with vendors (SUN, IBM, SAS, NetAPP, ECCS, HP) I can only be envious of someone who expects tech support from one. I can only recall a single instance when a vendor fixed something in response to my trouble report. Even in single vendor setups, the DB support group will have no hesitation to blame the hardware group, or the OS group for problems and close the problem report as 'solved'. If there is a vendor that doesn't do that, someone please mention them here. And I don't know any vendor that would not be perfectly comfortable closing a problem report with 'That is a known bug documented in TR123456. No patch or workaround is available'. Only a few favored customers actually get tech support, the rest might as well be buying at a flee market. The best way to protect yourself is to make sure that your configuration and purpose are very similar to a favored customer, or a target market.
The advantage Capture has in creating PDFs from OCR is that when Capture is not sure of an interpretation it just puts a little picture of the word into the PDF file, instead of taking a wild guess. The resulting file is searchable using ordinary PDF tools, on any words that were successfully OCRed, but of course the little pictures aren't indexed. We have found that Capture is a bit too optimistic and thinks it has the words down when it doesn't, and this limits our use of Capture, but in principle using a combination of OCR and pictures is the way to go.