Child pages
  • Tuple Space
Skip to end of metadata
Go to start of metadata


To provide, through a standard web-service interface, a simplified tuple-space-like capability which can be used by other services and by applications running in the ELF container.

We have chosen to implement this service instead of using JavaSpaces, the tuple-space which has long been part of the Jini Technology package (see and also, for the following reasons:

  1. There are serious problems with using multicast technologies on NCSA's production resource subnets.
  2. We do not require the full peer-to-peer capabilities offered by Jini.
  3. We do not require the full generality of JavaSpaces with respect to serializable classes; our service accepts an open-ended (i.e., basically untyped) wrapper object containing as its payload element a serialized XML blob (which usually, but not necessarily, will implement


Service WSDL


  • write takes a serialized string representing the ncsa.service.vizier.tspace.TspaceEntry object (see below); returns the entry id.
  • writeB is a batch method which does the same thing as write on multiple entries; returns array of ids corresponding to each entry.
  • read and take are batch methods, both requiring a serialized string representing the ncsa.service.vizier.tspace.TspaceMatchRequest object (see below), along with an integer indicating the maximum number of entries to return (default is 10; it is advisable not to set this number too high, as Axis will have trouble managing extremely large SOAP packets); the array returned contains serialized ncsa.service.vizier.tspace.TspaceEntry objects.
  • take also requires a timeout (if set to t <= 0, the non-blocking method is called), whereas takeNB will return immediately subsequent to issuing the request against the data store.
    • takeNB can return null if no entries satisfy the request constraints.
    • take will not return null unless timeout <= 0.
    • take may return an empty array (length 0) if its timeout has been exceeded.


The object stored in the service's data store and which can be retrieved through the read or take ports. A fuller description of the object is found on the page.


When passing this object to the write port, it is important that the id attribute/field be handled correctly.

  • If id is null, the object will be added as a new entry.
  • If id is not null, an attempt will be made to update an existing object with that id. The fields subject to update are:
    • timeTaken
    • timeCompleted
    • retried
    • ranOn

If the id is set, it is not necessary to provide a complete description of the entry; only the XML elements or attributes corresponding to the 4 updateable fields above will be processed. See further the page.


The object used to express the match query parameter required by the read or take ports. A fuller description of the object is found on the page.

  • No labels