In case a resultset is combined with other search terms, it doesn't matter how the server handles this, but in case one wants to request a specific record from a list (for example to get another format) it does matter whether the query is re-executed or that the record is obtained from the resultset, because you want exactly that specific record.
Perhaps it would be a good compromise to distinguish between two cases:
1) the resultset as a separate parameter in which case it refers to a specific set of records and is used instead of a search query
2) the resultset as part of the query in which case is doesn't matter how the server handles the query.
So either a query is specified or a resultset is specified (and the query may contain a resultsetid).
>>>> [log in to unmask] 31-05-02 15:49 >>>
>Theo van Veen wrote:
>> I assume that:
>> 1) resultSet in a boolean operation refers to the original query and not the actual list of records >(server will generally generate a new query)
>In the abstract it refers to the original query. From an implementation point of view it can be either (as >Ralph notes). The distinction between the
>query string and the result set is intentionally blurred. In early SRW meetings some argued against >even having a result set name. Eventually the
>people who wanted a result set name won that issue, but the distinction remains blurred.
>Thus a query string may be interpreted by a server to refer to a result set, a new query, or re-execute >the query. A result set name may be
>interpreted to refer either to a previously executed query string -- meaning either access the existing >result set (if it's still around) or
>re-execute the query -- or, the server may have discarded the query string and retained the result set >in which case the server would (obviously)
>interpret it to refer to the result set.
>> 2) resultSet stand alone in a query refers to the actual list of records (to be able to request specific >records from a resultset)
>The result set alone in a query has no special meaning, and is not distinguished from the case where >the entire query string is supplied instead. The
>server may choose to treat these two cases differently but they aren't different from the protocol point->of-view.
>> 3) the parameters startRecord and maximumRecords do not effect the resultset (the resultset refers >to the complete set)