Using the right version of MSXML in Internet Explorer

引自：http://blogs.msdn.com/xmlteam/archive/2006/10/23/using-the-right-version-of-msxml-in-internet-explorer.aspx

I’ve been working closely with the IE team leading up to the release of IE7 and looking at the use of XML in the browser.  During this investigation one thing has become immediately obvious – there is a lot of confusion around the versioning story for MSXML and how to instantiate the “right” MSXML object in the browser.  Here’s a quick snippet of code that I’ve seen all too often out on the web:
 
What NOT to do...
 
if (Web.Application.get_type() == Web.ApplicationType.InternetExplorer) {
        var progIDs = [ 'Msxml2.DOMDocument.6.0', 'Msxml2.DOMDocument.5.0', 'Msxml2.DOMDocument.4.0', 'Msxml2.DOMDocument.3.0', 'Msxml2.DOMDocument' ]; // MSXML5.0, MSXML4.0 and Msxml2.DOMDocument all have issues - be careful when using.  Details below.
       
        for (var i = 0; i < progIDs.length; i++) {
            try {
                var xmlDOM = new ActiveXObject(progIDs[i]);
                return xmlDOM;
            }
            catch (ex) {
            }
        }
       
        return null;
       }
 
The code iterates through the "progIDs" array and instantiates the highest version MSXML DOM that is available on the machine and returns it to the caller (see below for details on which versions ship where).  This has at least a few implications:
Compatibility – We do our best to maintain compatibility across versions of MSXML, however earlier versions of MSXML like MSXML 3 and MSXML 4 were implemented during the “wild west” of the XML emergence and we’ve learned a lot since then.  In addition, MSXML5 for Microsoft Office Applications was targeted specifically at Office scenarios.  Sometimes we do need to make design or implementation changes that affect behavior across MSXML versions.  By iterating through the versions of MSXML you open your app up to more potential risk of “finding” one of the differences unexpectedly.
Robustness - We can't fix every bug in every version so we've targeted MSXML6 (our newest) and MSXML3 (our most broadly deployed) versions of MSXML where we'll make the biggest investments.
Test cost – The more versions of MSXML your application potentially depends on means more versions to test your application with before you can ship it to your customers. 
My goal for this post is to give a quick history of MSXML lifecycle and versions, provide details with an example on implementing best practices with MSXML on the web, and talk about a couple key things to watch out for.
 
If you want the full story please read on, but if you’re short on time and want the quick scoop here it is in 4 bullets:
Use MSXML 6.0 - it is “in the box” on Vista and available for download on Win2k, XP, and 2003.  It has the best security, performance, reliability, and W3C conformance
MSXML 3.0 is our preferred “fallback” - It is installed on every OS from a fully patched Win2k SP4 installation on up, so it requires “zero-deployment” and is serviced regularly with the OS
MSXML 4.0 was released to the web about 5 years ago, but at this point has been superseded by MSXML 6.0 and is only intended to support legacy applications
MSXML 5.0 for Microsoft Office Applications is purpose-built for Office applications and isn’t intended for broad deployment.  Internet Explorer 7 actually has the MSXML5 components "off-by-default" in the Internet zone so your customers will get a goldbar for each MSXML5 control on a page if your code tries to instantiate it.  The best recommendation is to avoid MSXML5 in your web apps (only machines with Office 2003 or higher will have it, anyway.).
MSXML Lifecycle & History
 
OK, the full story requires a little bit more context – so let’s cover the different versions of MSXML, where they ship, and what the long term strategy is. 
 
Over the long run, the goal is to have our customers move their applications to MSXML6.  In terms of deployment, we want to ship our technology "in-the-box" with the operating system so that page authors and app developers can take advantage of it with zero deployment.  However, our customers have told us they want symmetrical XML APIs on all supported OS platforms, so we still need a way to get the newest XML technologies to our downlevel OSes (Win2k, Win XP, and Win 2k3). 
 
MSXML6 will be part of the Vista operating system when it releases, but requires a redistributable package to be installed downlevel.  We’d like to get MSXML6 “inlined” in the next service pack of each of the downlevel OSes, but we need a strong business case to do so.  So in the short and medium term we will continue to ship a redistributable package for MSXML6 that can be installed on downlevel operating systems.  We'll try to get a post up on the benefits of moving to MSXML6 sometime soon.
 
As much as we'd love everyone to be on MSXML6 today, we realize the migration can take some time.  So we're continuing to invest in MSXML3 to support existing applications and applications that have zero deployment requirements.  MSXML3 doesn't have all the improvements in MSXML6, but developers should consider it a robust and stable platform for MSXML applications.  MSXML3 is already part of the operating system on a fully patched Win2k SP4 installation and higher so in general no deployment to the client is required.  Going forward, MSXML3 updates will come out in each of the OS service packs.  MSXML3 SP7 is the last update to MSXML3 that should ship in redistributable form and in the future no redistributable package should be necessary for our partners and customers to use MSXML3 functionality.
 
MSXML4 was a predecessor to MSXML6 but hasn't ever shipped in the operating system.  MSXML6 is a significant step forward in terms of reliability, security, W3C and System.Xml compatibility, and it also has support for native 64-bit environments.  Right now we are investing much more heavily in MSXML6 and MSXML3 and we're encouraging our customers to move to 6 when possible and 3 when necessary.
 
Finally, anyone using MSXML5 who isn’t writing applications specifically targeted at Microsoft Office 2003 or Microsoft Office 2007 should migrate to MSXML6.
 
The details
 
Once you pick a version of MSXML to use, how do you do it effectively?  MSXML ships side-by-side with version dependent ProgIDs.  That means two things:
Versions are isolated – For example, if you’ve already got MSXML3 (msxml3.dll) on your machine and you install MSXML6 (msxml6.dll) it will lay down side-by-side in System32 and will have no effect on any application that uses MSXML3
ProgIDs are locked to their version - If you want your app to take advantage of your new MSXML6 installation you need to instantiate your MSXML objects using the MSXML 6.0 ProgIDs.
var xmlDOM = new ActiveXObject('Msxml2.DOMDocument.3.0') //uses MSXML 3.0
var xmlDOM = new ActiveXObject('Msxml2.DOMDocument.6.0') //uses MSXML 6.0
One related note - service packs of a particular version of MSXML are not side by side and will upgrade that version of MSXML to the service pack version.  For example, if your computer is running MSXML3 SP5 and you install MSXML3 SP7, all applications that use MSXML3 will automatically be upgraded to run on 3 SP7.
 
Ideally, customers should standardize on MSXML6, but as mentioned above legacy applications or zero-deployment requirements may block full migration to MSXML6 in the short run.  In this case there are two tensions that need to be balanced – functionality and test costs.  This essentially leads to two options:
Try MSXML6 and fallback to MSXML3 – MSXML6 has some improvements that aren’t in MSXML3 such as support for XSD schema, and improved stability, performance, and security.  So your app may try out MSXML6 if it is on the box and then “fallback” gracefully.  Just remember to test your app with MSXML6 and with MSXML3 so you aren’t surprised when you release your application.  Here’s a quick example:
if (Web.Application.get_type() == Web.ApplicationType.InternetExplorer) {
        var progIDs = [ 'Msxml2.DOMDocument.6.0',                                                   'Msxml2.DOMDocument.3.0'];
       
        for (var i = 0; i < progIDs.length; i++) {
            try {
                var xmlDOM = new ActiveXObject(progIDs[i]);
                return xmlDOM;
            }
            catch (ex) {
            }
        }
       
        return null;
       }
Standardize on MSXML3 with an eye towards MSXML6 in the future – This limits functionality somewhat to what is in MSXML3 but also keeps down test costs.  I’ll try to post something in the future about writing MSXML3 apps that should upgrade more easily to MSXML6 (and beyond).
A couple things to watch out for
MSXML6 has security sensitive features “off-by-default” while MSXML3 has some security-sensitive features “on-by-default” to avoid problems with backwards compatibility.  Check out the SDK for more details.
Use of XSD schema - MSXML3 does not have support for Xml Schema (XSD 1.0) so applications that depend on XSD will need to use MSXML6 directly.  There are a few changes from MSXML4 and MSXML5 in the XSD implementation in MSXML6 to be more conformant with the W3C specification and more compatible with System.Xml in .Net 2.0 so some apps may need to do a little work during upgrade.  See the SDK for more details.
Default Query Language - When you are querying the DOM with SelectNodes or SelectSingleNode the default selection language in MSXML6 is XPath while the default selection language in MSXML3 is XSLPatterns.  To switch MSXML3 to the standard XPath 1.0 query language set the second-level DOM property Selection Language like this - xmlDoc.setProperty("SelectionLanguage", "XPath"); see our SDK for more details.
Version Independent ProgIDs – There’s a lot of confusion around the “version-independent” ProgID for MSXML.  The version-independent ProgID is always bound to MSXML 3 (a lot of people think it picks up the latest MSXML that is on the box).  This means the version independent ProgID and the “3.0” ProgIDs will return the same object.  For example both statements in the following code will return an MSXML 3 DOMDocument:
var xmlDOM = new ActiveXObject('Msxml2.DOMDocument.3.0')
and
var xmlDOM = new ActiveXObject('Msxml2.DOMDocument')
Older ProgIDs - Stay away from ProgIDs that end in suffixes lower than “3.0”.  In particular some older operating systems have MSXML 2.6 on them, however these ProgIDs are “kill-bitted” in the recent MS06-061 Security patch
MSXML2 vs. Microsoft namespace – I’ve also seen a lot of code that instantiates the “Microsoft.XMLHTTP” ActiveX object rather than the MSXML2.XMLHTTP.3.0 or MSXML2.XMLHTTP.6.0 if you’re using 6.0.  The “Microsoft” namespace is actually older and is only implemented in MSXML3 for legacy support.  It’s unfortunate we used the “better” name on the older version, but stick to the “msxml2” namespace when instantiating objects.
As always, I'd love to get some feedback from people so if you have questions, comments, or random thoughts you'd like to share please comment here or mail me directly (email link above).
 
Adam Wiener
Lead Program Manager
Data Programmability/XML Technologies

收集的一些问题：
1. MSXML2.XMLHTTP.5.0的问题
http://hi.baidu.com/lontoo/blog/item/72d4091f665cb0cba6866950.html/cmtid/f0616f161b45114121a4e969
2. If you use MSXML 5, please stop
http://webdesign.about.com/b/2006/10/27/if-you-use-msxml-5-please-stop.htm
3. MSXML5...Not in This IE
http://blogs.msdn.com/ie/archive/2006/10/27/msxml5-not-in-this-ie.aspx