[GNUnet-SVN] r30515 - gnunet-java

[gnunet]

Author: dold
Date: 2013-11-05 11:19:43 +0100 (Tue, 05 Nov 2013)
New Revision: 30515

Modified:
   gnunet-java/ISSUES
Log:
issues

Modified: gnunet-java/ISSUES
===================================================================
--- gnunet-java/ISSUES  2013-11-05 09:57:06 UTC (rev 30514)
+++ gnunet-java/ISSUES  2013-11-05 10:19:43 UTC (rev 30515)
@@ -1,138 +1,39 @@
-there were still some [gnunetd] sections, should I delete them if I see one?
+set:
+ * generation collection and other stuff in set is implemented now
+  * some structural changes were necessary to implement everything
 
-gnunet-java now has EcdsaSignedMessage<M> and EddsaSignedMessage<M>,
-the design is a bit different from C GNUnet and replaces the 
E(c|d)dsaSignaturePurpose
+ * GNUNET_CONTAINER_HashMapIterator vs GNUNET_CONTAINER_MultiHashMapIterator
+  * very confusing name clash
 
-mesh: what config value do I have to tweak so that a newly started
-listening mesh handle receives a tunnel that has been created before?
- * is this even possible?
+ * there's now also an external iterator for multihashmap32
 
-I've implemented automatic tests in bash
- * somewhat elaborate bash hacking
-  * see cleanup (better alternatives?)
-  * killing the JVM properly (see gnunet-daemon-ballot-tally)
- * authorities have to be entered into the ballot on the fly
- * ... same with timestamps
- * ... and identities
- * some aspects are not possible to test due to crypto mismatch
- * bash tests aren't run by gradle automatically (yet)
+ * documentation: I'll meet with cfuchs on wednesday
 
-https://gnunet.org/bugs/view.php?id=3089
-The voters need the threshold encryption system's public key.
-How do they retrieve it?
-Asking authorities (which return a list of signatures from all other 
authorities)? 
-Is it the issuer's responsibility?
 
-voting tools:
- * should gnunet-ballot -r register with all authorities or one?
+doxygen:
+ * would it be possible to remove e.g. the call graphs to make
+   the doxygen on gnunet.org more accessible?
+  * see e.g. https://gnunet.org/doxygen/de/d22/crypto__hkdf_8c.html
 
-gnunet-config has no way to delete an option :-(
 
----------------------------------------
+secretsharing:
+ * discuss api
+ * how should shares be serialized
+  * ... and should they contain the list of peers?
+  * => maybe a configuration file / string?
+ * what if one of the authorities does not send some of the signed share parts 
s_{i,j}? who / how to blame?
+  * one idea is to use consensus with encrypted signeds{i,j}, so only the 
receiver can open it and
+    it can be known who sent their share parts 
+  * but: what if the encrypted value is wrong? there's no way to demonstrate 
this without revealing the
+    private key
+ * Should broadcast steps in the secret sharing protocol
+  be replaced by a consensus on signed commitments?
+  * this way all peers will have the same result
+ * what parameters should be the default?
+  * q (where q divides (p-1)) determines the size of the group, so how large 
do we want it?
+    (for practical purposes and security)
 
 
 
-transport:
- * documentation for StartMessage seems totally wrong, am I missing something?
 
 
-What now works:
- * consensus+testbed testcase in pure Java
- * group certificates and corresponding tool implemented
- * authorities now use consensus (though only tested with one authority)
- * ballot tool is now drastically simplified (one <400LOC file)
-
-
-gnunet-ballot-group-certify [OPTIONS]...
-Create a certificate attesting group membership for a given key.
-The resulting certificate file written to standard output.
-  -c --config=FILENAME         path of the configuration file
-  -h --help                    print this help message
-  -v --version                 print version
-  -L --log=LOGLEVEL            configure logging to use LOGLEVEL
-  -l --logfile=LOGFILE         configure logging to write logs to LOGFILE
-  -e --ego=EGONAME             name of the identity ego to use for signing
-  -m --member=KEY              key to certify membership for
-  -g --group=NAME              group to certify membership of key for
-  -x --expire=DATE             expiration date, in local time
-
-
-gnunet-ballot [OPTIONS]... BALLOT
-Create, modify and execute operation on ballots.
-  -c --config=FILENAME         path of the configuration file
-  -h --help                    print this help message
-  -v --version                 print version
-  -L --log=LOGLEVEL            configure logging to use LOGLEVEL
-  -l --logfile=LOGFILE         configure logging to write logs to LOGFILE
-  -e --ego                     ego to use for the operation
-  -q --query                   query election result
-  -s --submit                  submit the vote to the authorities
-  -r --register                register an election with the authorities (use 
with -e)
-  -i --issue                   sign the ballot as issuer (use with -e)
-  -x --select=CHOICE           select and encrypt a vote option (use with -e)
-  -V --verify                  verify signatures in the ballot and show 
information
-  -g --group                   incorporate the group cert into the ballot
-  -t --template                write a template ballot to the give ballot file
-
-
- * as the ballot tool is not an API, doesn't it make most sense to have
-   voting tests as shellscripts?
-
-YES.
-
- * how should FOREVER be handled in config files?
-
-There are keywords (NEVER, FOREVER, see util/strings.c).  I would probably
-for now not bother to check if the options are sane (i.e. vote finishes at 
eternity).
-
-
-Multiple authorities:
- * what should happen when duplicate vote is detected? where do we "complain"?
-   (can't always detect this on submission, as two auth. might get different 
votes)
-
-GNUNET_log (WARNING)
-
- * When is the threshold crypto set up? Don't we need another
-   time for when authorities start to set up the shared secret?
-
-True.
-
-
-exceptions and static initializers: when moving a class and not running 
'gradle msgtypes',
-this nice message pops up:
-
-java.lang.NoClassDefFoundError: Could not initialize class 
org.gnunet.construct.MessageLoader
-       at 
org.gnunet.construct.OptionalUnionTest.setupMessageMap(OptionalUnionTest.java:36)
-       at 
org.gnunet.construct.OptionalUnionTest.test_optional_union1(OptionalUnionTest.java:42)
-       at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
-       at 
sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:39)
-       at 
sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
-       at java.lang.reflect.Method.invoke(Method.java:597)
-       at 
org.junit.runners.model.FrameworkMethod$1.runReflectiveCall(FrameworkMethod.java:47)
-
-Any idea on how to improve the error reporting?
-
-Not really.   Just don't have those issues in a release ;-).
-
-
-question on testbed service disconnect adapter:
- * why is this needed? I get the idea, namely that the operation of service 
connect
-   is "canceled" (by calling done) and then the code for disconnecting is 
called, but why wouldn't
-   I want to do this myself?
-
-
-For symmetry, and because testbed may want to trigger OTHER operations that
-were delayed (due to limited resources) until the disconnect was done.
-
-
-* crypto mismatch is a bit annoying, when will libgcrypt be done?
-
-WK and I talked yesterday. He said this weekend, or earlier.  There were some
-additional minor design issues to be discussed first...
-
-
-consensus+testbed test looks *horrible*, do you have any suggestions on how to 
improve?
-
-If you have a consensus command-line tool, you can try shell scripting
-instead (and use gnunet-testbed-profiler to start multiple peers from
-the shell).