Hi Philipp, Adjusting re-transmission interval based on exponential backoff algorithm comes under WS-RM policy assertions. This means that each server/client can publish their own set of policies and they can exchange these information at runtime to decide weather they can successfullt communicate with each other. Since sandesha 1.0 (current version) is build on top of Axis 1.x and since Axis 1.x does not support WS-Policy their will be no affective advantage in having support for policies such as exponential backoff algorithm within sandesha, unless you use that for all your communications. However we will implement the support for exponentiona backoff for anybody who want to do it that way.
Actually WS-RM specification allows the servers to have their own limit on the highest message number they support. >From the spec: /wsrm:Sequence/wsrm:MessageNumber This required element MUST contain an unsignedLong representing the ordinal position of the message within a Sequence. Sequence MessageNumbers start at 1 and monotonically increase throughout the Sequence. If the message number exceeds the internal limitations of an RM Source or RM Destination or reaches the maximum value of an unsignedLong (18,446,744,073,709,551,615), the RM Source or Destination MUST issue a MessageNumberRollover fault. So we have not thought of going for the upper limit of unsigned long at this moment. Also using BigInteger objects may have a performance and memory usage drawback than using the long datatype. Thanx, Chamikara On 5/25/05, Philipp Reinecke <[EMAIL PROTECTED]> wrote: > Hello, > > I have some questions regarding Exponential Backoff as a > means to adjust the retransmission interval, and since I > did not find anything related on the web, I thought it > might be a good idea to ask some developers: > > (1) Exponential backoff as such only means lengthening of > the interval, however, common applications (e.g. TCP) em- > ploy methods of shortening it, too. How do you interpret > exponential backoff as given in the WSRM specification -- > in the strict sense or in a way that allows for a shrin- > king interval, too? > > (2) What is this adjustment meant to relate to? Should it > happen to individual messages or should it affect all mes- > sages in the same way? > > (3) This arose from my looking at the code: Sandesha has > some support for the exponential backoff strategy, but > there seems to be no actual adjustment of the interval. > Is this right? > > > As a follow-up to the old thread about the message number > problem in Sandesha: I did change my copy of the code to > use BigInteger instead of long, however, I then got dis- > tracted and Sandesha's development seems to have progres- > sed in the meantime. Would it be of use if I re-applied > the modifications to a more current version and sent a > patch, or has the problem been dealt with already? > > Regards, > > Philipp Reinecke >
