"Alex Krawarik [MSFT]" <alexkr@o...> wrote in message 
news:%23wHs9shSHHA.5068@T......
>
>>> So:
>>>
>>> Is this a known problem?
>>> Has anybody seen it before?
>>> Any good ideas on how to avoid it? (current plan is to simply issue a 
>>> GET if
>>> the time since the last request was about 10 seconds - the GET will 
>>> either
>>> force the new connection or keep the old one alive).
>
> Not known, no.

What would be the best way of making it known :-)

>>> Oh yes - this is using the XMLHTTP which you get with IE 6, on either 
>>> Win2K
>>> or WinXP. Code is running in javascript on a normal html page. Server is
>>> apache, but I don't think that's relevant. Code is this simple:
>
>>>  var httpObj = new ActiveXObject('Microsoft.XMLHTTP');
>>>  httpObj.open( "POST", url, false );
>>>  httpObj.setRequestHeader( "Content-Type", "text/plain" );
>>>  httpObj.send( postData );
>>>
>
>> Try MSXML2.XMLHTTP.3.0 and make sure the machine is fully patched.
>
> In general using the version-dependent progIDs is smart, because at least 
> then there is no ambuguity (for you, the developer) as to which version of 
> MSXML you are targetting. In this case Microsoft.XMLHTTP really does map 
> to MSXML2.XMLHTTP.3.0, so you're using the same object under the covers 
> (for supported OS configs).
>
> My first question would be...you are using XMLHTTP to send text/plain 
> data?

That was just a test harness. The real app is using XML both ways.

> Why? Just use the much-lighter-weight-and-still-safe-for-scripting 
> WinHTTP.WinHTTPRequest.5.1. If all you are doing is sending some text data 
> over the wire, and you are not doing XML processing with the HTTP 
> response, then don't use XMLHTTP!

Still, good to know about this.

cheers,
clive