presented. First of all, what performance and availability parameters are to be calculated and 

what are they to be calculated against? The end user of a service delivered via the Web 

views performance in terms of how fast the response to a request arrives and typically this 

means how soon the complete object appears in their browser. Of course, this is over 

simplifying the complexities involved in a HTTP transaction. Service functionality is usually 

delivered by unique URLs for Web delivered service and these URL will have different 

request/response rates depending on the amount of data passed to and from the server and 

the processing required on the server side. For example, the request may have an entity body 

associated with it as occurs with the submission of the contents of an HTML form to the 

server using a POST method. This data associated with the request entity may result in an 

increased delay in the complete request being received at the server. There is typically a 

delay associated with processing the request on the server and this may vary depending on 

what resource has being requested. The requested resource may involve some queries to a 

remote database for example and this may also significantly increase the processing time of 

the request on the server before the reply is sent to the client. Finally, the reply will usually 

contain content of some type and depending on the size, will again have an impact on when 

the complete reply is received at the client. 

This implies that: 

The amount of data sent with each request varies which may impact when the 

complete request is received at the server 

The processing time of a request at the server varies depending on the resource 

requested. The content of reply messages may be dynamically generated while 

others may be returning data representing static objects such as images on the 

server file system 

The amount of data sent with a reply can vary for a request, even for the same 

resource, in the case where it is dynamically generated. 

So the specification of performance parameters for services needs to flexible to take into 

account the circumstances of how the end user interfaces with a service(s) via HTTP. As 

was mentioned already, some derived parameters are more suitable to specific situations. For 

"