Page 2 of 9: First Prev Next Last
Web Log
Google And Ye Shall Find
By Andrew Price, 2008-01-30 23:51:59 in Geekdom. Leave a comment (Permalink)
I've just had an odd Google experience while scouring the web for relevant papers for my computer science project. It all started at the Google Scholar site which I was using to search for papers relating to Linux file systems and storage, particularly performance analysis, block I/O and anything to do with blktrace or GFS2.
After a while I stumbled across a highly relevant looking abstract of a paper entitled Evaluating Block-level Optimization through the IO Path [Riska, Larkby-Lahet & Riedel] so I clicked on the link. A username/password dialog appeared. It turned out I needed a USENIX account to get at this potentially juicy morsel of knowledge. So I clicked on the Web Search link that Google Scholar places next to search results, hoping to find the paper offered freely elsewhere. Expectedly, the first page of results showed links to pages on the USENIX site including a link to the actual PDF which again popped up a username/password dialog (Google's crawler must have a USENIX account...?).
Then I noticed something else - a View as HTML link. Surely it couldn't work. It shouldn't. Could Google search really be that nifty (or subversive)? I clicked on the link. Before my eyes the once-imprisoned-as-a-password-protected-PDF paper loaded in all its oddly formatted glory. A brief skim through reveals that some figures have missing information but I think I can live with that. I'll put it with the others and read it in the morning. Thanks, Google.
[Edit: Good news - since I wrote this entry USENIX has announced that it is making all the proceedings of its conferences available to everyone.]
Twyt 0.8 - Coming At You Like A Buzzard
By Andrew Price, 2008-01-27 14:01:40 in Twyt. Leave a comment (Permalink)
I've just released Twyt 0.8. It includes added support for Twitter user functions, better HTTP error handling, improved documentation and status messages posted with the twyt command line client now show up as 'from Twyt' instead of 'from web'. The Fedora and Debian python-twyt packages will be updated shortly.
[ P.S. This will be the first of my blog entries to pop up on Planet Fedora, so hello to all you Fedorable people! ]
Beer, Speech and Linux File Systems
By Andrew Price, 2008-01-12 11:58:36 in Geekdom. Leave a comment (Permalink)
Wakey, wakey. Rise and shine. -- My mother, somewhen between 1982 and 2008
It's been a while since I last wrote a blog entry that wasn't a software release announcement so I think it's time to remedy that. Since attending LinuxConf Europe (I enjoyed it thoroughly) which I mentioned in my last non-announcement blog entry life has been mostly revolving around university and the final year of my computer science degree, which has been a bit of a rollercoaster. But life is fun, in a fantastically geeky way.
The most interesting aspect has been my 3rd year project. The end product of which will be a file system performance analysis utility. Most pleasingly it will be an open source project, although I'm probably limited to accepting patches only once my dissertation is submitted. I won't bore you with the details but the project specification is similar to Chris Mason's Seekwatcher, which is the must-have accessory for the discerning Linux file system developer (darling).
The project is fairly challenging since it requires some specialist knowledge (I've probably spent 80% of my project time reading, 15% experimenting/exploring and 5% coding so far) and I've already learned a lot about file system implementations, aspects of the kernel and other interesting subjects that raise my state of mind closer to supreme enlightenment. The open source aspect of the project really does boost my enthusiasm for it and adds to my determination to roll out a good piece of software. That's the plan anyway.
The opportunity to do this project was presented to me by Steve Whitehouse of Red Hat GFS2 fame and I'm working under his guidance as a sort of "industry partnership" as the computer science department refers to it. Apparently Red Hat doesn't have an official student project scheme of this nature so I consider myself lucky that Steve has taken the initiative to offer the project and dedicate some of his time to answering my dumb questions. He's quite an inspirational character. I'd like to see these kinds of opportunities being offered to students by open source companies more often.
Here's to a fun 2008. May it be filled with inspiration, aspiration and perspiration.
Twyt (Reprise)
By Andrew Price, 2008-01-06 17:21:10 in Twyt. Leave a comment (Permalink)
Since the majority of twyt's code is reusable I've decided to brand it as an alternative Twitter API Python module instead of just a command line client. It's pretty simple to use, as exemplified by the following exciting and contrived Python session:
>>> from twyt import twitter, data
>>> t = twitter.Twitter()
>>> t.set_auth("username", "password")
>>> return_val = t.status_update("Test")
>>> s = data.Status()
>>> s.load_json(return_val)
>>> print s
[568644162] Andy Price: Test (Sun Jan 06 16:10:53 +0000 2008 via web)
>>> print s.id
568644162
>>> t.status_destroy(s.id)
A quick run through the code: set_auth() tells the Twitter object your username and password. Then status_update() sends your status message to Twitter and return_val is set to the JSON formatted string that Twitter sends back. We then load the JSON string into a Status object which lets us access the returned data (the details of the new status message) conveniently. We then pass the id of the new status message into status_destroy() which tells Twitter to delete the status message, because it's boring.
The docs in the classes are pretty good so help() should be your friend. Go forth and write super-shiny twyt-based applications.
Twyt is now in Fedora Rawhide and Debian Unstable, so it should propagate into Ubuntu Hardy Heron (universe) soon. Those of you brave enough to run Rawhide, Unstable and Hardy can install twyt using 'yum install python-twyt' or 'apt-get install python-twyt' respectively. Alternatively, just grab the latest tarball.
Twyt - A Command Line Twitter Client
By Andrew Price, 2007-12-27 05:53:53 in Twyt. Leave a comment (Permalink)
Allow me to introduce my new little python program named Twyt (randomly chosen but could be backronymed to "That's what you think").
Twyt allows you to post twitter messages, delete them, view your friends' latest messages and view other various twitter "timelines" from the command line. It's very young and really only does enough stuff to keep me happy so it might be lacking in features that others may desire. Patches welcome, as usual.
I wrote twyt a couple of weeks ago after I couldn't find a twitter client that met my standards and stumbled across the Twitter API docs. If you prefer GUI clients, it shouldn't be too hard to import twyt's twitter.py (or commands.py for that matter) and write a nice, shiny front-end for it. I've tried to document it quite well for that purpose - see the GUIHOWTO file for a possible example. Twyt uses the JSON interfaces to the Twitter functions and depends on the simplejson python library.
Have fun.
Pybackpack 0.5.4 - A File Is For Life...
By Andrew Price, 2007-12-24 16:20:06 in Pybackpack. Leave a comment (Permalink)
Blatantly copy&pasted from the mailing list:
Seems that people file bugs when I make releases so here's another one with some more bug fixes. I'm fresh out of wrapping paper and tinsel so .tar.gz will have to suffice :-)
Changelog:
New in v0.5.4 (23/12/2007)
--------------------------
+ Code cleanup based on pychecker results
+ Use genisoimage instead of mkisofs
+ Rearrange the source tree for easier installation and packaging
+ Set the backup set path on creation (bug #60)
+ Strip newlines from MRU entries
Note that the source tree has been rearranged a bit and pybackpack now depends on genisoimage instead of mkisofs so packaging will need to be updated.
Pybackpack 0.5.3 - Amidst The Storm
By Andrew Price, 2007-12-09 16:57:04 in Pybackpack. Leave a comment (Permalink)
Over the last week I've pushed out two versions of pybackpack. The first (0.5.2) included some background changes to encapsulate backup set data in two classes: BackupSet and BackupSets. This makes it easier to write extensions such as Seth Vidal's nautilus extension. I've updated his script and sent him the patch. The nifty highlight:
- total_files = backup['filelist_inc'] + real_files
- total_files.sort()
-
- flist = []
- for file in total_files:
- flist.append((file, True))
- extrakeys = {'default_dest':backup['default_dest'],
- 'removable':backup['removable']}
-
- rdiff_interface.WriteSet(backup['name'],
- backup['desc'],
- flist,
- True, extrakeys)
-
-
+ backup.files_include.extend(real_files)
+ backup.write()
Yay.
Oh, and 0.5.3 was a hurried bug-fix release because I'm an idiot.
LinuxConf Europe 2007
By Andrew Price, 2007-08-31 23:49:41 in Geekdom. Leave a comment (Permalink)
Tomorrow I'll be heading to Cambridge for LinuxConf Europe 2007 which starts on Sunday. It'll be my first open source related conference so I'm moderately excited and apprehensive.
It'll also be my first time in Cambridge which looks like a great old place with plenty of culture to soak up. I haven't signed up for the tutorials on Wednesday so I might play the tourist game before I return.
I'll have more to say once I get back. Perhaps a few photos too. See you there, if you're going.
Pybackpack Gets Some Nautilus Love
By Andrew Price, 2007-08-23 04:07:13 in Pybackpack. Leave a comment (Permalink)
Seth Vidal has knocked up a clever little script that lets you add files and directories to a backup set by simply right-clicking on them in a nautilus window. Thanks, Seth.
The pybackpack list is now aflutter with ideas for the future and ways to improve pybackpack's extensibility. Yay for rekindled enthusiasm.
Remember, kids, a backed up file is a happy file.
...The Person Got Sponsored
By Andrew Price, 2007-06-22 00:21:03 in Pybackpack. Leave a comment (Permalink)
Following on from my previous entry about getting sponsored to upload pybackpack into Fedora, I'm pleased to say that I've now been sponsored. I was surprised that it came so soon but after writing the previous blog entry and requesting a sponsor on fedora-devel-list@, Jeremy Katz got in touch with me and, after I had made a few necessary fixes to my packaging, approved it and sponsored me. I'm now awaiting the pybackpack CVS module to be created and then I can go ahead and maintain my package in Fedora. Happy days. Thanks to all the people who took time to review my RPM packaging and provide excellent feedback.
To Sponsor The Package Or The Person?
By Andrew Price, 2007-06-02 02:59:05 in Geekdom. Leave a comment (Permalink)
As you may know, pybackpack originally started off as a Google Summer of Code 2005 project for Fedora. Since I took over maintaining it, I've been keeping the Fedora review request alive in the hope that pybackpack would eventually get to where it was first intended. Of course, my distribution of choice is Ubuntu and most of my open source work that isn't pybackpack is Ubuntu-orientated so getting pybackpack into Fedora hasn't been high on my list of priorities, but I would like it to happen.
The first Fedora review request bug for pybackpack was opened in September 2005 and the latest bug (it had to be redone when extras was merged with core) has been open since January this year. Since then I've managed to get pybackpack sponsored into Debian (and consequently, Ubuntu). The Debian sponsorship process took hardly any time at all - just a day from ITP (intent to package) to upload, if I recall correctly. I thought it might be interesting to compare Debian and Fedora's new package processes point-for-point to see where the big difference lies:
Debian:
- Uploads of new packages are sponsored by Debian Developers
- The process:
- File an ITP bug
- Build your package
- Upload package to somewhere accessible
- Request a sponsor
- Fix any problems highlighted and repeat until approved
- Package uploaded by a sponsor
- Maintenance of your package is your responsibility, through the sponsor, who will encourage you to improve your packaging skills and start your journey to becoming a Debian Developer (if that's your goal).
- An alternative is one of Debian's teams (e.g. the python team or the games team) will take over maintenance of your package. This is similar to how the Ubuntu MOTU team works, and from what I've seen, it works very well.
Fedora:
- Package submitters are sponsored instead of packages, so they need to be granted access in order to upload their own packages.
- The process:
- Build your package
- Upload your packages somewhere accessible
- Request a review on the bugzilla
- Have informal reviewers comment on your package
- Make required fixes and repeat until approved
- Meanwhile, prove your knowledge of the process by reviewing others' packages or submitting more packages of good quality
- Wait for a sponsor to pick you up
- Get sponsored
- Apply for the access in the Fedora account system and finish up the rest of the process for becoming a contributor
- Upload your package using CVS
- Your packages are maintained by you (correct me if I'm wrong). By this point you've already been become a Fedora maintainer, since it was a prerequisite.
Looking at this comparison, I can't help but feel that Fedora has made it too difficult and daunting for new contributors to get new packages into their archive. Especially busy ones who are the upstream maintainer and have other projects to concentrate on as well. It's true that you get upload permission at the end of the process but not everybody wants to, or has the ability to, get that deeply involved. If Fedora wants to attract more packages they should probably decouple the process of becoming a Fedora contributor with upload rights, which is a lengthy process, from the process of getting a package approved for quality, which is a short and focussed process.
So, I guess I'll have to do some reviewing of other new packages and cut through more red tape before I get pybackpack into Fedora. I'm not totally against that, but my heart isn't really in it and I'd rather spend my time working towards becoming an Ubuntu MOTU and improving pybackpack. If anyone would like to take over maintaining the Fedora package of pybackpack from me, please do get in touch.
Pybackpack 0.5.1 - Towel Not Included
By Andrew Price, 2007-05-27 02:24:44 in Pybackpack. Leave a comment (Permalink)
Pybackpack 0.5.1 has been released. It contains a hefty portion of small tweaks to improve ease of packaging and fix a few bugs. See the changelog for more details. You can find the latest version of pybackpack on the downloads page on the wiki.
Pybackpack deb packages are now available in Debian's unstable and testing repositories and consequently, from the universe repository in Ubuntu Gutsy Gibbon. Packages for 0.5.1 should be uploaded soon.
Talk: FLOSSing Your Bits
By Andrew Price, 2007-05-16 04:47:51 in SUCS. Leave a comment (Permalink)
Yesterday the second round of the Swansea University Computer Society lightning talks was held. I gave a talk entitled FLOSSing Your Bits - a student's guide to starting an open source project. The slides are online [PDF] for anyone who might be interested in starting their own free/open source project. It's not a complete (or unbiased) guide but I think I've highlighted the important aspects. The slides should make some sense on their own but a video of the talk might be available online some time soon. Thanks again to the computer society for a very enjoyable and educational evening.
Happy 2nd Birthday, Blog
By Andrew Price, 2007-05-08 01:19:03 in Celebrations. Leave a comment (Permalink)
Today my blog is two years old. Aww, isn't it cute.
Since the 8th May 2005, I've managed to write a very humble 106 entries and attract a similarly humble 55 non-spam comments. I'd like to thank all three of my regular readers (my mother being one) and if any of you think I should make my blog more readable or worthwhile, do feel free to contribute a comment to bolster next year's statistics.
Reading back over my first few entries, I notice just how much I've learnt about the world, Linux and the tech world in general and I also notice how I've changed as a person. For instance, I rant a lot less these days. That's a good thing. One unfortunate change that seems to have struck me is that my writing style is a lot more formal and less quirky these days. I really should go back to writing like I did in such entries as "I Wish I Was a Naked Penguin" but perhaps to not such a degree of silliness.
Anyway, I'm starting to see why people keep journals and write blogs, now. It's cool to look back on yourself two years ago and see how you've progressed. Even though you write for the world to read, it's enjoyable to read the entries back for yourself, even if it does just teach you how immature you were back then.
One thing's for certain - I don't half ramble on sometimes :)
Ding Dong, Dell
By Andrew Price, 2007-05-02 04:50:19 in Ubuntu. Leave a comment (Permalink)
So, Ubuntu being chosen by Dell as the Linux distro to ship on some of its machines is old news now, but what does it mean for the future? I've had some thoughts about the advantages it could have in the bigger picture but for it to pan out that way there are some caveats which I'll explain later.
The first advantage might be the effect of inspiring other large PC retailers to offer some flavour of Linux with their machines. This would be a big advantage to us consumers as prices might become more competitive as companies vie for customer favour in that market. You might start to see Christmas time bargain commercials with shiny new computers flying across the screen with the Ubuntu logo clearly in the corner of the desktop on daytime television.
It could also be a poke in the ribs for hardware vendors. If large PC retailers start to find that buying hardware which has good Linux support or open specifications/drivers is advantageous to their business models, more hardware vendors might wake up and open source their own specifications and drivers in order to attract those companies to their product lines. Dell might even see fit to fight in the corner of open source software and lobby hardware vendors to go open source just because, say, Michael Dell agrees with the free software philosophy (whether he does or not, I don't know).
Organisations with existing Dell contracts might benefit. I'm not entirely sure how these things work, but here's an example based in reality. My university seems to be keen on Dell PCs and I'm guessing that to buy that many machines they would need to work out a big contract deal with Dell which I'm also guessing would have some kind of time frame attached to it. If the university had to choose from the range of Dell products, and they came across Ubuntu, or maybe even an Edubuntu option for academic organisations (wink), they might recognise the added benefits of using open source software solutions in their networks and opt for that instead. I'm sure my university would prefer an operating system which didn't propagate worms, viruses and other malware throughout its network. Who wouldn't?
The one effect I'm sure about is that the news will raise the profile of Ubuntu and Linux as a feasible alternative to Windows. People browsing Dell's product range will see Ubuntu as an operating system option (maybe even a lower price tag, we'll see) and look into what advantages it would hold for them (free upgrades forever, etc.) and compare it to Windows before they've tried it. It'll help to put "Ubuntu" into the vocabulary of a wider audience and for people who tried Linux back in the 20th century and gave up on it because it wasn't user-friendly enough, it'll help to update their perceptions of Linux.
Of course, I did say there were a few caveats and here they are. In order for the Dell news to have these desired network effects, the whole industry would need to sit back and watch how Dell's business is doing in relation to it. Back in 1999, Dell tried to market Linux as a desktop solution but then decided there wasn't enough demand to carry on. This time, I feel success for Dell's Linux desktop range is more likely but as a realist I'm not entirely convinced. Microsoft has a large amount of people locked into its software and that lock-in is very strong. I think, in the Average Jill's mind, the switch to Linux will need to be sparked off by something greater than a few dollars difference in PC prices. They'll need to take a leap into the virtually-unknown after all. I hope Dell will market their new offering enthusiastically and wisely and in doing so help to carry over the whole feeling behind using Ubuntu and other open source solutions.
My prediction about the hardware vendors taking notice is highly optimistic. Ever since I tried to get an old PC's modem working with Linux back around 2000 I've known about the stubbornness of hardware vendors and their bad habit of keeping their driver code buried as if it was a bloody dagger with their fingerprints all over it. It takes a very courageous CEO of a hardware company to push the decision to go open source (or open specification), and I applaud any that do. I doubt the recent news will directly push any hardware companies into going open but it'd be a nice surprise if it did. As Mark Shuttleworth mentioned on the video interview about Ubuntu and Dell, when Linux became a bigger player in the server market, the hardware vendors took note and decided to start becoming more Linux friendly in the context of that market. I hope they'll do the same for the desktop market.
It's a hard one to predict, but this could certainly turn out to be a big turning point in the life of open source software and I'm excited about the future. I hope as many people who told Dell that they'd buy a PC with Linux pre-installed go out and buy one, to keep our end of the deal.