Currently the core requires the module file, protocol and role to be specified, which results in a local protocol being generated in a module named "".

Although this may be useful in situations where only the individual protocol is required - other situations (such as in tooling environments that operate on the module level), would be easier if the project could be done based on the module and role - resulting in a module named "_" and containing a project of all the protocols in the global module.

In scribble 0.3 spec it introduces the concept of a module path, similar to the Java concept of a class path, to provide zero or more directories that can be searched to find a module.

Not sure yet whether this should be done, as the resulting repository will be very small. However it would also be targetted to producing a particular type of tool.

If we have already decided to separate out Eclipse and eventual Web App tooling, then having separate CLI repo makes sense - but there again it could be the default approach bundled with the core project.

Enable travis CI.

The initial requirement is to support simulation of a trace (series of sends/receives) against a set of FSMs - one per role (or service) being simulated.

However the more general case will be to manage a set of session instances, per role/service, over time - with the events being received in realtime. It is also possible that there will be multiple instances of the monitor (for load balancing/resilience purposes), so the session data needs to be persisted in a shared db/cache (should be pluggable to allow different implementations).

This means that the session needs a way to save the position within the FSM, for a particular session instance (and service/role), and then be able to recover it when a new event for an existing session instance is received.

So in the case of OpenTracing - the role will be determined by the service name, and the session instance by the trace id.

Consider separating CL scribblec scripts aimed at Scribble built by Eclipse vs mvn install.

Although the only difference (currently) between them would be the classpath entries for each. Perhaps factor out just that part, so that changes to other parts of the script do not need to be repeated for both.

Currrently, the single script looks for both Eclipse and mvn build outout on a single classpath, with higher priority for the Eclipse output. This works fine if one or the other is not actually present.

Have a launch configuration type (with Run As menu support) that can operate on .trace files.

Experiment with creating junit xml files (in a temp area?) and opening the editor on it to show the results.

Creating this issue as a placeholder for discussions about multicast and broadcast communications.

In scribble we can specify multiple 'to' roles which would enable different endpoints, representing different behaviours, to receive and use the same message.

However wondering whether we also need to cater for multiple (potentially unbounded) instances of a role(s) - I think we would need to find some real world examples to understand whether this is relevant.

If defining a global protocol, then would we have multiple recipients all behaving the same way - but doing something differently internally, to warrant it being considered equivalent from an externally observable behaviour perspective?

I am an undergraduate CS student, as part of my honours thesis I have been studying MPST and trying to find use cases for Scribble. I've been trying to follow the tutorial at http://www.scribble.org/docs/scribble-java.html, and have been trying to execute the Adder implementations provided.

I apologize that I am still new to Java and maven so I'm probably just not doing something very simple. If anyone can help it would be appreciated.

When I try to compile, for instance, AdderS.java, I get the following error:

scribble-java-master\scribble-demos\scrib\tutorial\src\tutorial\adder\AdderS.java:16: error: package tutorial.adder.Adder.Adder does not exist
import static tutorial.adder.Adder.Adder.Adder.Add;
                                        ^

Which makes me think that I need to compile the file tutorial/adder/Adder/Adder/Adder.java

Attempting this gives the error

scribble-java-master\scribble-demos\scrib\tutorial\src\tutorial\adder\Adder\Adder\Adder.java:7: error: package org.scribble.core.type.name does not exist
import org.scribble.core.type.name.Role;
                                  ^

Am I going about this the wrong way? Have I missed some part of setting up the software? Or am I not supposed to try to compile these files this way? I am using Windows 10 and trying to run the software on the command line. I apologize again for my inexperience.

See if possible to upload the distribution zip as part of the release on github.

Reference that may help: https://gist.github.com/stefanbuck/ce788fee19ab6eb0b4447a85fc99f447

The wiki documentation does not seem to line up with the code.

https://github.com/scribble/scribble-java/wiki/SJDGProtocol references classes like org.scribble.parser.ProtocolParser pp=new org.scribble.parser.ProtocolParser(); which seem to no longer exist since v0.4.x

should we not be using this version 4? is there any documentation for v4 that describes how to parse scribble syntax as it seems to have been rewritten or at least heavily refactored from the from the current developer guides

I am trying to follow the tutorial at http://www.scribble.org/docs/scribble-java.html
I used eclipse to "Maven install" the project and extracted the resulting zip file.
On Windows, both scribblec.sh --help and scribblec.sh scribble-demos/scrib/tutorial/src/tutorial/adder/Adder.scr -api Adder C -d scribble-demos/scrib/tutorial/src/ have the same result: a black window with a caret opens and immediately closes. No endpoints were generated for the adder.
The window has the following title: "usr/bin/bash --login -i" followed by the path of scribblec.sh and the parameters (e.g. --help).

Microservices tend to use common infrastructure patterns to make the systems more resilient, for example circuit breakers to prevent services from being overloaded by returning a 'default' response, retry mechanisms, fallbacks, etc.

We need to examine how these are handled from a scribble protocol description and monitoring perspective.

When a new version is tagged, trigger an automatic release mechanism.

I noticed that scribblec generates package paths with capitalized name if I use capitalized protocol names, i.e. if I name my global protocol "Sample" under module org.example with file name scribble, the Java package path will be inconsistent with Java naming convensions: org.example.scribble.Sample is the package name. But it seems like in Scribble doc, capitalized names are preferred as they are also used as part of class name, and the naming convention for Java classname is indeed Capitalized. Is there a way to generate API packages that are both compliant with Scribble and with Java conventions?

Need to check that checkstyle plugin is being used and is correctly configured.

Main example is message, where a parameter name or type alias could be used, which may also reference a type imported from another module.

Need some utility that can be given the context within a description, and also use cached modules (so not to continually reload on each request), to return the actual type being used.

If a local model is not consistent with the behaviour of the associated global model, then offer a quickfix in the Eclipse tooling to update the local model in the file.

When compiling using mvn clean && mvn install with JDK 12.0.1, the scribblec.sh file after unzipping scribble-dist/target/ does not load a classpath for scribble-ast, which results in the following Exception when executing:

Error: Unable to initialize main class org.scribble.cli.CommandLine
Caused by: java.lang.NoClassDefFoundError: org/scribble/ast/ProtoDecl

Manually adding scribble-ast classpath into scribblec.sh fixes this issue:

 99 ARGS=
100 
101 CLASSPATH=$DIR'/scribble-cli/target/classes/'
102 CLASSPATH=$CLASSPATH':'$DIR'/scribble-core/target/classes'
103 CLASSPATH=$CLASSPATH':'$DIR'/scribble-ast/target/classes'
104 CLASSPATH=$CLASSPATH':'$DIR'/scribble-lang/target/classes'
105 CLASSPATH=$CLASSPATH':'$DIR'/scribble-parser/target/classes'
106 CLASSPATH=$CLASSPATH':'$DIR'/scribble-codegen/target/classes'
107 CLASSPATH=$CLASSPATH':'$DIR'/scribble-main/target/classes'
108 CLASSPATH=$CLASSPATH':'$ANTLR
109 CLASSPATH=$CLASSPATH':'$DIR'/'$LIB'/antlr.jar'
110 CLASSPATH=$CLASSPATH':'$DIR'/'$LIB'/antlr-runtime.jar'
111 CLASSPATH=$CLASSPATH':'$DIR'/'$LIB'/commons-io.jar'
112 CLASSPATH=$CLASSPATH':'$DIR'/'$LIB'/scribble-core.jar'
113 CLASSPATH=$CLASSPATH':'$DIR'/'$LIB'/scribble-lang.jar'
114 CLASSPATH=$CLASSPATH':'$DIR'/'$LIB'/scribble-parser.jar'
115 CLASSPATH=$CLASSPATH':'$DIR'/'$LIB'/scribble-codegen.jar'
116 CLASSPATH=$CLASSPATH':'$DIR'/'$LIB'/scribble-main.jar'
117 CLASSPATH=$CLASSPATH':'$DIR'/'$LIB'/scribble-cli.jar'
118 CLASSPATH=$CLASSPATH':'$DIR'/'$LIB'/stringtemplate.jar'
119 CLASSPATH="'"`fixpath "$CLASSPATH"`"'"
120 
121 usage=0

Errors displayed to users should be subject to i18n.

Currently the CLI is not implemented - and the only test ignored.

Need to update to include support for command line usage.

Enable an automatic code test coverage mechanism.

See whether possible to separate out code into modules associated with model, parsing, projection, well-formedness etc.

Due to role not having positional information, updated error messages to be reported against GMessageTransfer - however if the role does not exist, the highlighted position is the message signature (-1).

Need to investigate how the positional information is assigned to the model objects.

Also change trace extension to ".trace", and include the test number as a suffix on the module name, e.g. [email protected].

This will allow the .trace extension files to be used in automated testing.

MessageTransfer validation rule has been changed to associate errors with the complete element, until positional information is available on the role.

Currently the DefaultParserContext is only receiving these roles as a token, rather than ModelObject, so not applying the properties - need to investigate.

As raised by @rhu1 via email, it would be better for the artifact folder names to reflect the artifact id - and as the modules folder no longer has its own pom (as the artifacts are in the main org.scribble groupId), it would be better to locate them under the top level folder.

Therefore,the sub-folders in the existing modules folder should be moved up a level (removing the modules folder), and renamed to include prefix scribble-. The root pom.xml should then be update to reference the new location/names.

Need examples where nested choice paths have same start behaviour that gets merged.

Provide support for projecting a global model into local models per role.

Need to examine some concrete scenarios of how timeouts are handled in a microservices context.

Ray's code currently uses exceptions for reporting issues found in a Scribble definition.

However when integrating into Eclipse or other tooling (e.g. webapp) we will want to provide a list of errors and allow the user to navigate to the appropriate part of the scribble definition associated with the error.

Project the global model to the local model associated with the role of the local model being validated, and then ensure that the local models are consistent.