Jeremy Allison on ‘The elephant in the room – free software and microsoft’. While he works at Google, this talk was ‘off the leash’ – not about Google . As usual – grab the  video We should care about Microsoft because Microsoft’s business model depends on a monopoly [the desktop]. Microsoft are very interested in ‘Open Source’ – Apache, MIT, BSD licenced software – the GPL is intolerable. Jeremy models Microsoft as a collection of warring tribes that hate each other… e.g. Word vs Excel.

The first attack was on protocols – make the protocols more complex and sophisticated. MS have done this on Kerberos, DCE/RPC, HTTP, and higher up the stack via MSIE rendering modes, ActiveX plugins, Silverlight…  The EU case was brought about this in the ‘Workgroup Server Market’. MS were fined 1 Billion Euros and forced to document their proprietary protocols.

OOXML showed up rampant corruption in the ISO Standards process – but it got through even though it was a battle against nearly everyone! On the good side it resulted into an investigation into MS dominance in file formats -> MS implemented ODF and MS have had to document their old formats.

MS have an ongoing battle in the world wide web – IE / Firefox, ajax applications/ silverlight.

All of these things are long term failures for MS… so what next?… Patents . Patents are GPL incompatible, but fine with BSD/MIT. The Tom Tom is the first direct attack using MS’s patent portfolio. This undermines all the outreach work done by the MS Open Source team – which Jeremy tells us are true believers in open source, trying to change MS from the inside. Look for MS pushing RAND patented standards: such things lock us out.

Netbooks are identified as a key point for MS to fight on – lose that and the desktop position is massively weakened.

We should:

• Keep creating free software and content *under a copyleft license*.
• Keep pressure on Governments and organisations to adopt open standards and investigate monopolies.
• Lobby against software patents.
• Search for prior art on relevant patents and destroy them.
• Working for a corporation is a moral choice: respectfully call out MS employees.

Jonathan Oxer spoke about the google Moon X-prize and the lunarnumbat.org project – it needs contributors: software and hardware hackers, arduino/beagleboard/[M]JPEG2000 gooks, code testers and reviewers, web coding, documentation, math heads & RF hackers. Sounds like fun… now to find time!

Paul McKenney did another RCU talk – and as always it was interesting… Optimisation Gone Bad (RCU in Linux 1993-2008). Linux 2.6 -rt patch made RCU much much much more complex with atomic operations, memory barriers, frequent cache misses, and since then it was slowly being whittled back, but there is now a new simpler RCU based around the concept of doing the accounting during context switches & tracking running tasks.

Glyn Moody – Hackers at the end of the world. Rebel code is now 10 years old… 50+ interviews over a year – and could be considered an archaeology now   I probably haven’t down the keynote justice – it was excellent but high density – you should watch it online

Glyn talks about open access – various examples like the public library of science (and how the scientific magazine business made 30%-40% profit margins. The Human Genome Project & the ‘Bermuda Principles’: public submimssion of annotated sequences. In 2000 Celera were going to patent the entire human genome. Jim Kent spent 3 weeks writing a program to join together the sequenced fragments on a 100 PC 800Mhz Pentium processor.  This was put into the public domain on just before Celera completed their processing – and by that action Celera were prevented from patenting *us*.

Openness as a concept is increasing within the scientific community – open access to result, open data, open science (the full process). An interesting aspect to it is ‘open notebook science’ – daily writeups, not peer reviewed: ‘release early, release often’ for science.

Amazingly, Project Gutenberg started in 1971!

Glyn ties together the scientific culture (all science is open to some degree) and artistic culture (artists share and build on /reference each others work) by talking about a lag between free software and free content worlds. In 1999 Larry Lessig setup ‘Copyright’s Commons’ built around an idea of ‘counter-copyright’ – copyleft for non-code. This didn’t really fly, and Creative Commons was setup 2 years later.

Wikipedia and newer sharing concepts like twitter/facebook etc are discussed. But… what about the real world: transparency and governments, or companies? They are opening up.

However, data release !=  control release. And there are challenges we all need to face:

• GFinancialC “my gain is your loss”. Very opaque system.
• GEnvironmentalC “my gain is our loss”

Glyn argues we need a different approach to economic governance: the commons. 2009 Nobel laureate for Economic Sciences – Elinor Ostrom – work on commons and their management via user associations… which is what we do in open source!

Awesome!

Pandora-build. There for support – I’ve contributed patches. Pandora is a set of additional glue and layers to improve autotools and make it easier to work with things like gettext and gnulib, turn on better build flags and so forth. If you’re using autotools its well worth watching this talk – or hop on #drizzle and chat to mtaylor

The open source database survey talk from Selena was really interesting – a useful way of categorising databases and a list of what db’s turned up in what category. E.g. high availability,community development model etc. Key takeaway: there is no one-true-db.

I gave my subunit talk in the early afternoon, reasonably well received I think, though I wish I had been less sick last week: I would have loved to have made the talk more polished.

Ceph seems to be coming along gangbusters. Really think it would be great to use for our bzr hosting backend. 0.19 will stablise the disk format! However we might not be willing to risk btrfs yet

Next up, the worst inventions ever.. catch it live if you can!

Another must-grab-the-video talk : Mako’s keynote. Antifeatures, principles vs pragmatism do come together. The principled side – RMS & the FSF – important to control ones technology because its important to control ones life. The pragmatic side – quality, no vendor lock etc. False dichotomy.. freedom imparts pragmatic benefits even though it doesn’t intrinsically import quality, good design:  95% of projects 5 contributors; median number of contributors 1, and such small collaborations are no different than a closed source one.

Definition of antifeatures – built functionality to make a product do something one does not want it to do. Great example of phone books: spammers pay for access to the lists, and thus we have to pay *not to be listed*, but its actually harder to list and print our numbers in the first place. Mako makes a lovely analogy to the mafia there. Similarly with Sony charging 50 dollars not to install trialware on windows laptops in the past.

Cameras: Canon cameras disabled RAW saving…. CHDK, an open source addon for the camera outputs RAW again. Panasonic are locking down their cameras to reject third party batteries.

The tivo is an example of how focusing on licensing can miss the big picture: free stack, but still locked into a subscription to get an ongoing revenue stream.

Dongles! Mako claimed there wasn’t a facebook appreciation group for dongles… there is.

Github: paying for the billing model – lots of code there to figure out how many projects in a repo, so that they can charge on that basis.

DRM is the ‘mother of all antifeatures’ – 10K people writing DRM code that no users want!

Gabriella Colemans keynote was really good; grab it from the videos once they come online.

WETA run Ubuntu for their render farm: 3700 machines, 35000 cores, 7kw per ‘cold’ rack and 22kw per ‘hot’ rack. (Hot racks are rendering, cold racks are storage). Wow. Another talk well worth watching if you are at all interested in the issues related to running large numbers of very active machines in a small space.

And a classic thing from the samba4 talk at the start of the afternoon: MS AD domain controllers do no validation of updates from other domain controllers: classic crunchy surface security. (Discovered by samba4 borking AD while testing r/w replica mode).

Blue-ray on linux is getting there, however one sad thing is that the Blue ray standard has no requirement that vendors make players be able to play un-encrypted content – and there are some hints that in fact licences may require them to not play un-encrypted content.

Peter Chubb’s talk on Articulate was excellent for music geeks: midi that sounds like music from lillypond.

Ben Balbo talked about ‘Roll your own dropbox’. Ben works at a multimedia agency, but the staff work locally and don’t use the file server…. use instant messenger to send files around! Tried using subversion… too hard. Dropbox looked good but 3-7 hundred a month – too pricey given an existing 1.4TB of spare capacity.

He then considered svn + cron but deleted directories cause havoc & something automatic was wanted… so git + cron instead. Key thing used in doing this was having a work area with absolutely no metadata. Conflicts dealt with by filename.conflict.DATESTAMP.HOSTNAME.origextention

Doesn’t trigger of inotify, no status bar widget, only single user etc at the moment, but was written to meet the office needs so is sufficient. Interestingly he hadn’t looked at e.g. iFolder.

For a while now I’ve been using subunit as part of my regular development workflow. I would pipe test results to a file, use subunit to report on failures from that file, and be able to inspect all the failures at my leisure without rerunning tests or copy and pasting from far back in my history.

However this is a bit adhoc, and its not trivial to get good pipelines together – while its not hard, its not obvious either. And commands like tee are less readily available for Windows users.

So during my holidays I started a small project to automate this workflow. I didn’t get all that much done due to a combination of travel and coming down with a nasty bug near the end of my holidays – which I’m now recovering from. Yay health returning + medicines. If only we had medichines .

However, I managed to get a reasonable first release out the door this evening. Grab it from launchpad or pypi.

Testrepository has a few deps – all listed in INSTALL.txt. Folk on Ubuntu Lucid should be able to just apt-get them all (sudo apt-get install subunit will be enough to run testrepository). If you’re not on Lucid you can grab the debs manually, or use the subunit ppa (sudo add-apt-repository ppa:subunit), though I’ve noticed just today that that karmic subunit build there only works with python 2.5, not the default of 2.6 – I will fix that at some point.

Using Testrepository is easy if you are developing python code:

$ testr init
$ python -m subunit.run project.tests.test_suite | testr load
id: 0 tests: 114

This will report any failures that occur. To see them again:

$ testr last
id: 0 tests: 114

The actual subunit streams are stored in .testrepository in sequentially numbered files (for now at least). So its very easy to get at them (for instance, subunit-stats < .testrepository/12).

If you are not using python, you can still use subunit easily if you are using shunit, ‘check’ or ‘cppunit’. subunit ships with bindings for shunit and cppunit, and check uses libsubunit with the CK_SUBUNIT output mode. TAP users can use tap2subunit to get a subunit stream from a TAP based testsuite.

It’s still early days but I’m finding this much nicer than the adhoc subunit management I was doing before.

Evolution recently moved to a sqlite summary db rather than a custom summary db implementation. Its great to see such reuse of code.

However, it’s not really a complete transition yet as I’ve had cause to find out today. I’ve blogged before about performance with the sqlite summary sqlite database. Today I was greet with a crash-on-startup bug which happily has a patch upstream already. Before I looked in the bug tracker though, I did some house cleaning.

I started with a 900MB folders.db. Doing a vacuum on the db dropped that to 300MB. It doesn’t appear to be something that evolution does itself. Firefox too appears to lack an automatic vacuum. sqlite is an embedded database, and its wonderful at doing that, but its not as install-and-forget as (say) PostgreSQL which does autovacuum. So an additional tip is vacuum your folders, e.g. with http://www.gnome.org/~sragavan/evolution-rebuild-summarydb, a helper script that will run vacuum on all your account summary db’s. Note that it *does not rebuild*, it solely vacuums, and as such does not add or delete (modulo bugs in sqlite) data to the summary database.

After the housecleaning, I checked that the sqlite database was in good condition:

sqlite3 folders.db
pragma integrity_check;

This returned a number of indexing issues, so I reindexed:

reindex;

Evolution now starts up and crashes in a fraction of a second - a big improvement. Finally, I started looking at the evolution code as I now was fairly confident it was a bug - it was in a sqlite callback function - and the column the function extracts data from (flags) is missing a NOT NULL constraint, but the code doesn't check for NULL - boom. From there to finding the bug report and existing patch was trivial.

And this is where my comment on reliability turns up: Evolution doesn't anticipate NULL flag values in its code, so why does it insert them into the database at all ? I suspect its due to some aspect of the incremental conversion to using sqlite summaries. More concerning for me is the possibility that there are many other such crash bugs lurking in the new sqlite based code.

There are possibly some clues as to the excessive table scans done by evolution in the use of a flags bitset rather than separate columns, but I haven't looked close enough to really say.

the emperor has new clothes

bzr has just changed the base class for its test suite from ‘unittest.TestCase’ to ‘testtools.TestCase’. This change has cleaned up a bunch of test logic, deleted a significant amount of code (much of which was redundant with Python unittest) and added some useful and important features.

bzr  has only been able to make this change due to testtools expanding its mission from a simple ‘aggregation of proven unittest extensions’ into one where new extensions that *make unittest more extensible*. My deepest thanks to Jonathan for permitting me to use testtools as the vehicle to put these extension-enabling-extensions (and for his patience in reviewing said changes!).

The change was pretty easy: The bulk of the changes were in bzrlib.tests and bzrlib.tests.test_selftest. I chose to cleanup an ugly API at the same time which added a little scattershot across a number of tests. And there are more changes that can be done to take better advantage of testtools – the amount of deleted and cleaned up code isn’t complete. Even so, its a pretty clear win:

18 files changed, 228 insertions(+), 496 deletions(-)

What went?

bzr had an implementation of TestCase.run. This function is the main workhorse of Python’s unittest module, and yet sadly it has to be replaced to change the exceptions that can be raised(to signal new outcomes), or to improve on test cleanup. Testtools provides an API to permit registering new exception types and handlers for them. Like python 2.7 testtools also provides the TestCase.addCleanup API, and these two things combined mean that bzr no longer needs to reimplement the run method.

For expected failures, bzr uses a helper method TestCase.expectFailure to perform an existing assertion and convert the test into an expected failure if that assertion does not trigger. This was another feature testtools already provides and thus got deleted.

All the custom code for skipping and expected failures got deleted, and the other outcomes bzr uses turned into extensions (as per the run discussion above).

In bzr test cases generate a log (because bzr generates a log) and previously the TestResult in bzrlib inspected each test that had been executed to extract the log. This was made simpler by using the details API that testtools provides (see testtools.TestCase.addDetail), which permits tests to add arbitrary data in a semi-structured fashion. This is supported by subunit and a long standing bug with bzr selftest --parallel was fixed as a result – logs from tests run in other processes are now carried across the process barrier intact and are presented cleanly.

Some other minor cleanups are in unittest compatibility code, where bzr would degrade gracefully with unittest runners, and testtools provides such logic comprehensively, so all that got deleted too.

Whats new?

I think the most significant new facility that testtools offers bzrlib is assertThat. This assertion is inspired by the very nice assertThat in JUnit (which has changed substantially since Python’s unittest was written based on it). This assertion separates the two concerns of ‘raise an exception’ and ‘decide if an exception should be raised’. The separation allows for better reuse of custom checking code, because it permits composition in a cleaner way than extra assertion methods permit. Testtools does not include many matchers as yet, but matchers are easy to write, and if one were to write a small adapter to the hamcrest library, there are a bunch of ready made matchers there (though they have a very Java feel – such as is not meaning is – which is why Testtools did not use that library).

Secondly, the addDetail API referenced above, in combination with testtools.TestCase.addOnException will permit capturing the entire working area when a test fails, something that developers currently have to fiddle about with breakpoints to achieve. This hasn’t been done, but is a straight forward patch I hope to do in the new year.

Lastly, Testtools offers testtools.TestCase.getUniqueInteger and testtools.TestCase.getUniqueString, which are not as yet used in bzr tests, but we may start using them soon.

Beyond that, the other features of testtools are already present in bzrlib, and we simply need to find and delete more duplicated code.

Recently I’ve been working on the Python unittest API in my spare time, with a long term goal of making it possible to safely and sensibly glue many different plugins together into the core.

Two important components of that goal are being able to extend the data included in a test result, and being able to change how a test is run (such as adding new exceptions that should be treated as specific outcomes – python unittest uses exceptions to signal outcomes).

In testtools 0.9.2 we have an answer to both those issues. I’m really happy with the data included in outcomes API, ‘TestCase.addDetail’. The API for extending outcomes works, but only addresses part of that issue for now.

Subunit 0.0.4, which is available for older Ubuntu releases in the Subunit releases PPA now, and mostly built on Debian (so it will propogate through to Lucid in due course) has support for the addDetail API. Subunit now depends on testtools, reducing the non-protocol related code and generally making things simpler.

Using those two together, bzr’s parallelised test suite has been improved as well, allowing it to include the log file for tests run in separate processes (previously it was silently discarded). The branch to do this will be merged soon, its just waiting on some sysadmin love to get these new versions into its merge-test environment. This change also provides complete capturing of the log when users want to supply a subunit log containing failed tests. The python code to do this is pretty simple:

def setUp(self):
    super(TestCase, self).setUp()
    self.addDetail("log", content.Content(content.ContentType("text", "plain",
        {"charset": "utf8"}), lambda:[self._get_log(keep_log_file=True)]))

I’ve made a couple of point releases to python-junitxml recently, fixing some minor bugs. I need to figure out how to add the extra data that addDetails permits  to the xml output. I suspect its a strict superset and so I’ll have to filter stuff down. If anyone knows about similar extensions done to junit’s XML format before, please leave a comment

Bzr build-deb is very nice, but it can be very tricky to get started. I recently did a fresh debianisation of a project that is in bzr upstream, and I thought I’d record the recipe to make it work (at least until the various bugs making it hard re fixed).

Assuming that the upstream uses bzr, it goes like this:

1. Start with a branch that is close to the code you want to Debianise. E.g. if the release was off trunk, 3 commits back: bzr branch trunk -r -3 debian
2. Debianise as normal: put the tarball with the right name in the parent dir,  add a debian directory and fiddle until you build a package you’re happy with. Don’t commit while doing this.
3. Build a source package- debuild -S, or bzr builddeb -S
4. Revert your changes – bzr revert.
5. Import the dsc – bzr import-dsc ../*.dsc
6. Now, you may find that some dot files, such as .bzrignore have been discarded inappropriately (there is a bug open on this). If that happened, keep going. Otherwise, you’re done: you can now use merge-upstream on future upstream releases, and debcommit etc.
7. bzr uncommit
8. bzr revert .bzrignore (and any other files that you want to get back)
9. debcommit
10. All done, see point  6 for details.

Hope-this-helps