FaeLLe wrote
Yes i figured it was Birelli's reference counting algorithm.
yes, remember that we followed Sun's spec, and it says on section 3.3: "RMI uses a referencecounting garbage collection algorithm similar to Modula-3's Network Objects. (See "Network Objects" by Birrell, Nelson, and Owicki)"
Our university research group has currently identified faults with that speific algorithm and I am working on an implementation of the revised version.
yes, we are aware of some of the faults, like creating a circular reference. BTW: Sun's implementation has the same problem.
However I was wondering if you could point me towards some literature/material that discusses these specific features that are imposed by RMI's dgc api spec ? These would be very beneficial.
Unfortunatelly we did not have any document describing these specific features; we worked based on Birrel algorithm and we implemented the API, while doing that we found some non-solved issues, and we created our own solution for them. What we did document were some issues we found on the spec. (follow this links for further info: http://www.itc.unc.edu.ar/javadev/rmi/knownissues.html#10 http://www.itc.unc.edu.ar/javadev/rmi/specissues.html#8 ) Daniel On 3/23/06, Daniel Gandara <[EMAIL PROTECTED]> wrote:
We use a "reference counting" algorithm based on the Modula3, plus specific features imposed by the rmi dgc's api spec. Daniel ----- Original Message ----- From: "FaeLLe" <[EMAIL PROTECTED]> To: <harmony-dev@incubator.apache.org> Sent: Wednesday, March 22, 2006 4:34 PM Subject: Re: ITC: Contribution of java.rmi Hello Daniel, Can you please tell us which algorithm you made use of for the Distributed Garbage Collector ? I am currently researching in this field and details would benifit me. Regards, - Vikram On 3/22/06, Daniel Gandara <[EMAIL PROTECTED]> wrote: > > Hi, > > We are thinking in making a release of java.rmi package compatible with > harmony's VM and classlib code (1.4.2), using the compiler options that > allow partial 5.0 language features, and removing all j.u.c classes we > use. > We know that we will loose performance, but it looks like it is worth a > try. > What do you think? > > Daniel > > PS: Faxes have been re-sent, please let me know if there is any further > problem. > > ----- Original Message ----- > From: "Geir Magnusson Jr" <[EMAIL PROTECTED]> > To: <harmony-dev@incubator.apache.org> > Sent: Sunday, March 19, 2006 10:45 AM > Subject: Re: ITC: Contribution of java.rmi > > > > Thanks - this is great. > > > > Please not that every faxed document that both I and Jim, the ASF > > Secretary, received was unreadable. > > > > I can't know if it was the original copy, or the fax machine, but we > need > > to have everything resent again. > > > > THanks > > > > geir > > > > > > [EMAIL PROTECTED] wrote: > >> As announced before we are contributing java.rmi package on behalf of > >> ITC > >> (Cordoba Institute of Technology). > >> > >> Below is a short description of the contribution. > >> > >> Daniel Gandara > >> > >> -------------------------------------------------------------- > >> > >> Package name java.rmi > >> > >> Package Description > >> > >> Clean room implementation of java.rmi following J2SE 1.5 spec > >> www.fitc.unc.edu.ar/javadev/rmi/index.html > >> > >> > >> Current Status > >> Currently our package contains implementation for: > >> - java.rmi > >> - java.rmi.dgc > >> - java.rmi.registry > >> - java.rmi.server > >> > >> package java.rmi.activation and deprecated method/classes > >> have not been implemented. Activation have been designed but > >> has not been implemented yet. > >> > >> Tools like rmic and rmiregistry have not been implemented. > >> > >> > >> Architectural Design > >> Package has been architected following a layered design, which > >> seem to be the most appropriate model for the RMI implementation > >> in accordance with the specification. > >> For further details check > >> www.fitc.unc.edu.ar/javadev/rmi/architecture.html > >> > >> > >> Javadoc > >> > >> Javadoc of the package can be reached here > >> www.fitc.unc.edu.ar/javadev/rmi/javadocRMI/index.html > >> > >> Testing > >> Unit and integration tests (and their documentation) are > >> provided with the code. > >> > >> > >> Implementation Notes > >> The code uses J2SE 5.0 features, such as generics, so it > >> requires 5.0 VM and libraries (i.e.: java.util.concurrent). > >> It has been tested against Sun SDK, removing the original > >> java.rmi.* and replacing it by ours. > >> www.fitc.unc.edu.ar/javadev/rmi/ideas_uro.html > >> > >> > >> Known Issues > >> www.fitc.unc.edu.ar/javadev/rmi/knownissues.html > >> > >> Specification Issues > >> www.fitc.unc.edu.ar/javadev/rmi/specissues.html > >> > >> > >> > >> > >> > > -- www.FaeLLe.com www.VikramMohan.com
-- www.FaeLLe.com www.VikramMohan.com