Spotlight on BIDS

Clive Massey, in his own words, [8th March 1996] reponds to the comments in the review of BIDS on the http protocol/persistent objects within the server.
There is a response from Jan Cropper to the response below.

I may have misunderstood Isobel's point about the http protocol, but it seemed to me to imply that persistent objects are not possible in the server and we think that this is not true.

The fact that http is stateless doesn't mean that the interface as a whole can have no concept of a state or of a session. There are 3 problems here (I think):

• the need to maintain a pseudo-session at the host (ie bids) end in order to keep the service reasonably secure, by enabling logout, etc. the server needs to be able to validate the username and password in some way, and then to be able to certify that an incoming connection is from a previously validated source. The most likely solution to this is for the server to supply a key with the form which can be validated when the form is resubmitted (by the user). This key would also contain a date & time stamp which would enable the server to time-out a series of forms in the same way that a telnet session is timed-out by the host if the user simply walks away without explicitly logging out.
• The need to maintain persistent objects in the server to enable the implementation of (eg) a marked list facility. persistent objects could be handled by passing the persistent data out to the browser every time as hidden fields in the html, but I think this is not a good solution when the objects are potentially large. Given that the server will have to maintain persistent data for session management, it seems to me to make good sense to generalise this mechanism to allow many sorts of objects to be "remembered". The solution is not elegant, but we think it is manageable.
• The need to maintain an instantiation of the database engine between connections, both to avoid having to reopen the databases each time the client connects and to allow the reuse of previous result sets (eg by ANDing them together) which you can currently do with the bids interface. this is more complicated. At the moment I'm not absolutely sure that it can be done. If it were to prove possible on unix, but not on vms, then this would be somewhat discomfiting for us. Even if it is possible, the fact that html forms make it much easier to specify searches which apply to several different fields simultaneously (eg "find me refs which have fruit flies in the title AND are written by jones OR by bloggs") may mean that the facility to recycle sets is not so important.

Contents Page - Electronic Libraries Programme and Project Information
Search Ariadne - Editorial for this issue - News desk - Reviews - Front Cover