This Technet page has more information including some important details about making sure to reset some of your policies before replacing the ADM files, as you won’t be able to edit them afterwards:

If you have previously configured any of the Group Policy settings affected by this update, you must set those policy settings to their Not Configured state before you remove the previous 2007 Office system ADM files and load the updated version 3 ADM files. This removes the registry key information for the policy setting from the registry. This is because if an .adm file is removed, the settings that correspond to the .adm file do not appear in Group Policy Object Editor; however, the policy settings that are configured from the .adm file remain in the Registry.pol file and continue to apply to the appropriate target client or user. This also applies to any policy settings that you had previously configured that are listed in “Removed settings” later in this article.

You can download the Administrative Templates and OCT in a self-extracting exe file. Included are ADM, ADMX and ADML files in various languages (English, French, Spanish, German, Italian, Japanese, Korean and a couple of flavours of Chinese).

Also has the OPA files and a settings reference, but this other page claims that this is the definitive version of the Office 2007 GP and settings file. I can’t tell the difference – they are the same size and have the same number of rows on the list pages, and have identical MD5 checksums, so they are the same file.

I suspect this was a newer version than the old version in the old download before the newer version superseded the old version so it is now the current version. Clear as mud?

Anyway, most of the focus of these is on fixing a few broken things and targeting settings relating to Open Document format files (making it the default for saving, or blocking it being used at all, that sort of thing.)

Happy policy making!

Overall I like the way the Office exams work – the real application (minus the help!) running in the top half, and the questions at the bottom. Each question has a few tasks to complete, and you are measured on the end result, not how you got there.

This is a much better test of real-world ability to use the software than any multiple-choice questions can ever hope to be. Yes, it means that you could take a few wrong turns, and click on some irrelevant buttons before finding the thing you were looking for, but you can do that in real life too. The exam is limited to 50 minutes, so you can only afford to do this on a handful of questions, and you need to be able to make up the time on other questions by reading it once and going straight to the correct feature or function.

For the MOS 2003 exams I had a few minutes to spare for every subject except Excel where I had to rush the last question in the last few seconds. This time round I finished all of them early, and Excel had the fewest questions and left me the most remaining time -19 minutes out of the 50 allowed, the rest between 11 and 15. The Outlook exam had a few technical hitches: in three scenarios, things which were referred to in the questions simply did not exist in the mock environment provided, but using the “reset” feature in each case reloaded and everything was fixed. Frustrating, but at least the reset button did the trick, so bear this in mind if you run into similar problems.

The Word, Excel and PowerPoint exams had several questions using the new formatting tools – style galleries, themes, colour schemes were all covered, as you would expect since these are the backbone of the new approach to making richer, more visual documents. MS also want to make sure you got the hang of using the options for making documents more secure and collaborating with colleagues – make sure you know how to prepare a document by stripping out metadata, protect a sheet/book in Excel, mark a file as final, manage comments and track changes. Outlook did not really have any surprises, and covered the whole breadth of email, contacts, calendaring, tasks, and the options to customise the environment (things like signatures and email format).

I felt all the Office exams were a little bit biased towards the new features, when you consider the overall breadth of the applications. As an upgrade from MOS 2000 or 2003 this would make sense, but someone taking a Microsoft application exam for the first time might be frustrated when they are not asked anything to test their knowledge of lots of other core areas. I only had two very simple tasks to do with charts, for example, which is hardly stretching their capabilities (of course you could get 20 in-depth chart questions, since every exam is different). At least I had some questions on Excel formulas this time, for the MOS exam I had none at all – which just shows how varied the exams can be, I suppose.

MCAS Vista is a slightly different exam format, using a simulation environment closer in style to the MCP technical exams than the Office ones. This means that you often can’t do things in many different ways, which allows you to quickly realise when you have tried an incorrect option as you get an error saying that feature is not available in the simulation. The simulation takes a few seconds to reload in between questions, but the clock stops during this period. Again, the best way to get through the exam in the time will always be to know the subject thoroughly so you can go straight to the right answer.

On the Vista exam there were also some straight multiple choice questions, which were generally quick to complete, but I guess they represent fewer marks than the tasks which have several steps to them. A couple of questions nearly caught me out as they referred to functionality which would be very different in a domain environment than in a workgroup network. Just be aware that this exam is geared more towards end users, home enthusiasts and people in smaller businesses who look after the systems in addition to their main job. If you are an IT support professional, then this could still be a suitable qualification, but it is more likely that the Vista Technical Specialist (MCTS exam 70-620) will be more relevant to your needs.

Right, now I just need to update the Office 2007 training and profile pages on my website and LinkedIn to show my five new MCAS qualifications!

This also has the bonus of including the Office Customisation Tool (OCT) which you can use to create an MSP file to customise a centralised network installation of Office for new installations, upgrades, or reconfiguration. You can find out more about the methods for customising Office 2007 setup files here and specifics about the OCT here. In addition the download extracts an Excel workbook “Office2007GroupPolicyAndOCTSettings.xls” that provides information about the 2007 Office release Group Policy settings and OPA settings, making it clear what can be pre-customised at the point of installation and what can only be set through policies.

You will probably also find the Office 2007 settings reference file useful. This is a comprehensive reference for all the settings in the GUI for Access, Excel, Outlook, PowerPoint and Word 2007. This gives the equivalent UI path in 2003 (where there is one), the default setting, what choices can be made, what policy settings exist and which registry keys those change. A very helpful file for understanding how to customise the user experience, and deciding which parts to do through policies and which settings are better left to users (and perhaps prompting you to educate them about the usefulness of some of these).

According to the article on the Excel team blog about this bug:

The first example that we heard about was =77.1*850, but it became clear from our testing as well as additional reports that this was just one instance where Excel 2007 would return a value of 100,000 instead of 65,535.  The majority of these additional reports were focused on multiplication (ex. =5.1*12850; =10.2*6425; =20.4*3212.5 ), but our testing showed that this really didn’t have anything do to with multiplication – it manifested itself with many but not all calculations in Excel that should have resulted in 65,535 (=65535*1 and =16383.75*4 worked for instance).  Further testing showed a similar phenomenon with 65,536 as well.  This issue only exists in Excel 2007, not previous versions.
…
Said another way, =850*77.1 will display an incorrect value, but if you then multiply the result by 2, you will get the correct answer (i.e. if A1 contains “=850*77.1”, and A2 contains “=A1*2”, A2 will return the correct answer of 131,070).

So, it is important to note that most calculations which result in numbers near to or equal to 65,535 and 65,536 will be absolutely fine. It is only through some very specific oddities about how floating point numbers work that you will get one of the 12 situations where this bug occurs. If it does, you will have cells that read “100,000″ rather than the correct answer. Anything else in Excel which you base on those cells will be correct.

You can add to them, multiply by them, show conditional formats such as colour scales or icon sets, even draw charts with those values and Excel will correctly handle the real, underlying value and not the displayed one. Macros or external programmatic methods of retrieving the cell’s contents also return the true stored value.

When does this actually give you the wrong values (rather than just a wrong display)?

Rounding the result will change to the incorrect value permanently

Some people find it confusing that when they have a column of figures added up, the sum at the bottom may not seem to be correct.

A typical example would be a column of prices including added sales tax (by multiplying by some decimal factor). These prices will only be shown to two decimal places in most currencies, but there will be a bit of rounding up or down to achieve this.

The sum of the underlying values may therefore differ from a sum of the displayed values, which gives the impression that there is a miscalculation somewhere, as shown in the example to the right. One way to avoid this is to use one of the ROUND functions to actually ROUND off the stored value to the nearest 1/100th (or force it to always ROUNDUP or ROUNDDOWN). The advantage of this is that if you change your mind you can easily change the formula to increase or decrease the accuracy of the underlying numbers, and my own recommendation would usually be to use this method.

If you ROUND a cell or calculation which has a buggy result you will get the wrong value as a fixed end result.
So, 2 * ROUND(850 * 77.1, 0) gives 200,000 as a result, whereas 2 * 850 * 77.1 gives the correct 131,070. This is true whether you use ROUND directly on a buggy calculation or on a cell containing the end result of one. Curiously, ROUNDUP and ROUNDDOWN are able to ignore the bug, and both ROUNDDOWN(850 * 77.1, 0) and ROUNDUP(850 * 77.1, 0) correctly result in 65,535 (since this is what the value “actually” is if you ignore floating point errors).

Telling Excel to use precision as displayed will keep the wrong value.

The alternative to the ROUND function to get over sums which appear inconsistent is to change one of your Excel options to “Set precision as displayed”. This means that if you change the number of decimal places shown in a cell (using the toolbar buttons or changing the number formatting) the actual stored value will change immediately and permanently to be rounded to that many decimal places. If you decrease the decimals to zero, then increase them again, all your numbers will be .0000… Of course, if the cell contains the result of a formula, you lose nothing, but if it contains data values they will be irrevocably changed. The point of this feature is to save you using those ROUND functions, and to make sure that what you see always appears to be calculated consistently.

It is set on a per-workbook basis, so it is perfectly possible for you to receive a workbook which already has this turned on. By the same token, if you use this feature you may actually confuse someone else who works on that same workbook, or they may permanently lose data accuracy by changing the formatting of cells if they are not aware that this is taking place.

In order to be able to see if it is turned on or not, and change it if you wish, in Excel 2007 go to the Office button at the top left, choose Excel options, then the advance section and look for the subsection labelled “When calculating this workbook” where you will find “Set precision as displayed” (as shown on the left).

For Excel 2003 the same option is in Tools > Options > Calculation under Workbook options and called simply “precision as displayed”. NOTE: the bug described in this post does not affect Excel 2003, the way to turn on or off this feature is only included here for completeness since the state of the feature is saved with the workbook, not as part of your Excel options.

If you have this feature turned on and your workbook contains calculation which have this bug, then those results will become permanently wrong (as 100,000 or 100,001). Any other calculations based on those cells will also be incorrect. Unless you can be sure to check every single cell in a workbook, which is only practical for very small files, you really need to turn this feature off when using Excel 2007 until a fix for the bug is applied.

Exporting incorrect values

As long as you handle the numbers within Excel, they work properly and simply display incorrectly. However, if you copy cells with these incorrect values and paste out to a text editor (such as Notepad), the incorrect display value is carried, because this always passes the number as shown, rather than the internal value. If you are only showing figures rounded to two decimal places (but not actually ROUNDing them) then Notepad would show the same as the sheet, and the underlying extra accuracy is lost.

While Notepad is useful to quickly see these bugs in action, it is not usually used as an import and export routine for large volumes of multi-column data. A common method is to export to a CSV file which can then be imported into another program (often a database which cannot read and write to Excel files directly). CSV files will also carry the incorrect display figure instead of the underlying stored value. This is crucial for anyone using Excel as a number crunching machine to export results to a third party application.

I have a client who uses Excel to churn out management reports about their product sales, profits and so on. The reports that people read in this case are summaries (often from pivot tables), and the chance of them actually seeing a number on those summaries with this bug are vanishingly small. However, they use the underlying basic calculation data to export and feed into their ERP system, and here there is a risk that a number could be passed through at the wrong value via a CSV file. Chances? Small, but not sufficiently small to be ignored.

FUD is in the air

There is unfortunately a lot of sidetracking and “fear, uncertainty and doubt” (FUD) in the comments to that article about floating point accuracy in principle. No-one has ever claimed that Excel or any other calculation program using floating point arithmetic is always totally accurate to an infinite number of decimal places, this is inherently unachievable using binary notation to store fractions and display the results in decimal. At face value it seems annoying that Excel does not treat the calculation of “simple” numbers such as 4.1-4 correctly and get exactly 0.1 (rather than 0.0999999999999996 which it does get).

The reality is that in most cases no-one is looking at enough decimal places to notice or care, or even for the error to compound up into something important. On the flip side, Excel will happily handle arbitrarily small binary fractions such as 1/18,446,744,073,709,551,616 + 1/4,294,967,296 without blinking. A human trying to work on such numbers would inevitably at some point convert them to that strange human notation called decimal and lose some accuracy.

Normally, Excel handles these rounding errors gracefully, and still displays something very near to the correct value, since the error is usually down in the dozens of decimal places. The oddity on this occasion is that for some reason it makes a bad assumption as to what constitutes a reasonable value to display and gets it rather visibly wrong.

I will of course post an update once a fix is available.