commit 3350885b7d3c24795b5ccf9e1fbeee379ebcecd0 Author: Isis Lovecruft isis@torproject.org Date: Sat Nov 3 01:19:53 2012 +0000
* Updated the TODO file. PLEASE READ IT. --- TODO | 110 ++++++++++++++++++++++++++++++++++++++++++++++++++++++------------ 1 files changed, 90 insertions(+), 20 deletions(-)
diff --git a/TODO b/TODO index 2686ef7..63d950c 100644 --- a/TODO +++ b/TODO @@ -1,26 +1,12 @@ This is a list of things to be done on ooni-probe.
-Once you have completed something you should add a -note to this file stating what you have done under -the item. +Once you have completed something you should add a brief note to this file +stating what you have done under the item. If you discover needed tasks, feel +free to add them, but also keep in mind that OONI is mostly using the Tor Trac +instance, and the main ticket for OONI which all tests should be organized +under is here:
-Migrate code from old ---------------------- - -Migrate all the interesting parts of the old code to the new. - -It's important to make the new code asych and based on Twisted. -It should respect the design goals of the new ooni-probe model. - -New things to develop ---------------------- - -These are either components specific to the new refactor of ooni -or that we haven't yet figured out how they should work. - -* Design and implement the Node Factory - -* Design and implement the Network Node and the Code Exec node classes + https://trac.torproject.org/projects/tor/ticket/5869
New things to test ------------------ @@ -36,3 +22,87 @@ New things to test nowhere is this presented to someone trying to run a test. So, the informing users/testers bit can be worked on, and the testing. Obviously we're going to want something more robust that a 20 LOC Makefile pretty fast. + +Finalization of API design +-------------------------- + +* The nettest.TestCase should have an interface. + + I know that there is a push away from using zope.interfaces, but I think + it is actually *highly* necessary for ensuring that subclasses implement + the required functions, and also that they do not improperly override + necessary functions, for them to run. + + Personally, I am quite annoyed when I subclass a class from Twisted and + override a public method, and it breaks things (when nothing in their + documentation informed me that it would break things) and I have to spend + half an hour digging through their code to figure out precisely what is + needed externally from the function I'm overriding. Others should not have + to do this with our code. + +* The nettest.TestCase should have a twisted.python.usage.Options subclass and + interface as well, even if the instantiation of that subclass is handled by + the ooni.oonicli or the ooni.runner. There is more functionality to + usage.Options that we should expose than merely "optParameters", for + instance the "coerceOptions" parameter validation methods, or the + "postOptions" configuration. + +New things to develop +--------------------- + +These are either components specific to the new refactor of ooni +or that we haven't yet figured out how they should work. + +* Finish implementing the backend collection code. + + o PCAP READER/WRITER: + This should be quite simple...see scapy.all.wrpcap and + scapy.all.rdpcap. However, we have been warned by other projects that + this does *not* scale well. For example, see: + https://github.com/isislovecruft/switzerland/blob/master/switzerland/client/... + Which is a circular ring buffer specifically for libpcap, to avoid kernel + buffer overflows due to a high number of incoming packets. I expect this + to only be an issue on substantially high-bandwidth nodes...though that + is what we'll be dealing with when we deploy on Mlab. + + o PCAP UPLOADER: + This also sounds simple, and is, until you begin to deal with things like + persistence. What we really need is rsync, written in python, or at least + some cross-platform implementation. I (Isis speaking) am the current + maintainer of pyrsync, BUT DO NOT USE PYRSYNC. It is only an + implementation of the rsync *algorithm* for diffs, it is not rsync the + program. Also, it is BROKEN AND I DO NOT MAINTAIN IT. If you want to + maintain it, please take it off my hands. + +* Useability: + + o UNITTESTS. Pronto. + + o DOCUMENTATION. If you found something that confused you, or still + confuses you, and you couldn't find the answer within fifteen seconds, + then that thing is not well documented. Make it better, or at least mark + it with an "XXX document me!" tag. + +* Persistence: + + o We need some type of scheduler/cron thing which will background the tests + so that they don't take up a terminal, and can be configured to run + certain tests at timed intervals. + + o The Reporter will probably need to be updated to handle knowing when *a + test* has completed, but that the scheduler is still running. + +Migrate code from old +--------------------- + +Migrate all the interesting parts of the old code to the new. This is mostly +finished, but there still are things in the /old-to-be-ported directory which +might be of use. At this point, because we have gone through several version +of the API design, many of them are entirely unusable, and merely the general +idea remains. + +It's important to make the new code asych and based on Twisted. It should +respect the design goals of the new ooni-probe model. Also, importing new, +non-standard libraries should be discussed first, if the new test is to be +used in the core of OONI (packaging scapy and twisted already makes our +codebase quite large). \ No newline at end of file