Missing LIFT once again…aaarrghh

February 7, 2007

For the second time in a row, I’m missing the LIFT conference, due to collisions with teaching gigs that were planned way too early.

I guess I’ll mark the dates in red for next year’s conference as soon as they are published.

Too bad…especially as the list of participants includes a lot of people that I’d haved loved to meet there. Next time.

The eBay architecture

December 19, 2006

Ugo points to the slides of a recent SAMSIG presentation on eBay’s architecture. A fascinating read…and it’s comforting to learn that they threw away most of J2EE: eBay scales on servlets and a rewritten connection pool.

The Venice Project opens up

November 16, 2006

tvp-image.jpgSeveral of the cool people who have been working on the super secret Venice project are finally talking about it publically Just follow the links.

Reading between the lines: Open Source rocks. And Open Source developers even more…

Behold, Internet Explorer 7 is upon us

October 19, 2006

The M$ download page (no I’m not linking to that;-) says upgrade with confidence. According to my colleagues who have tested it, this might indicate that IE7 faithfully reproduces most bugs of its predecessors, while adding a new set of fun and original ones. There’s a better way.

Worse, it seems like windows update will push this new and improved pile of…software to masses of unsuspecting customers. Be ready for some of these how come I suddently cannot read your site anymore? messages.

Google Docs PDF – powered by OpenOffice

October 12, 2006

Funny…a PDF exported from Google Docs says

PDF Producer: OpenOffice.org 2.0

I was hoping they’d use FOP – but at least it looks like they aren’t reinventing the wheel.

It’d be interesting to check if they’re using OpenOffice to generate other formats as well.

Google Docs – a browser is all you need

October 12, 2006

Google Docs combines text documents and spreadsheets in a unified interface, with PDF export for decent printable versions.

Editing is versioned, and collaboration features are built in – so this is geared towards getting things done, as opposed to spending hours polishing paragraph spacing and figthing with page breaks. Sweet.

This is a nice step towards getting rid of bloated computers and software and having just a browser to cover most people’s needs. The network is really becoming the computer.

Going to Venice

October 5, 2006

So this starts being more tangible.

Amazon Elastic Compute Cloud will change the way we look at servers

August 26, 2006

Upload your root Linux filesystem image to the Amazon Elastic Compute Cloud, turn it on and hey presto – you’ve got a new, reliable, pay-as-you-go server on the Web, with solid connectivity I assume.

I haven’t tried the service yet (the limited beta slots are currently full), but, if I understand correctly, customers pay only for the actual CPU power used. This would bring a server running at 20% average CPU load to about $15 a month (744 hours * $0.10 per instance-hour * 20%). Not including storage and bandwidth, but these are cheap for mainstream applications (update: the 20% factor does not apply, see below).

The virtual hardware is said to be predictably equivalent to a 1.7Ghz Xeon CPU, 1.75GB of RAM, so the above estimate would allow you to run interesting stuff already.

Virtual server instances can be duplicated at will (up to 20 without making special arrangements, currently) to handle more load, and decommissionned easily when they are not needed anymore.

I’m going to try this thing as soon as I get a beta slot. The implications for the kind of medium-sized systems that I’m involved with could be huge, in terms of flexibility, reliability and costs.

Update: according to the Amazon Web Services Blog post about EC2, CPU time is billed by clock hour, not by actual usage. Too bad, this makes the service more suitable for applications that actually need CPU power. But the price is still quite good for what you get.

Tim Bray on Atom

August 8, 2006

This morning while flying to work I listened to Tim Bray’s ETech 2006 presentation on Atom. Very interesting! Tim outlines several known problems with RSS, and how Atom avoids them.

Phil Windley has written a short summary of the talk, but it’s worth listening to, if only for the insights into how Atom came to life.

Atom is not just a feed format, and with the Atom Publishing Protocol it makes a very interesting API for content management and retrieval.

“Building scalable web sites” by Cal Henderson

August 3, 2006

I’ve just finished reading building scalable web sites by Cal Henderson the architect of Flickr. And, if you’re building web applications, I’d suggest that you do the same, it’s a great and very useful read.

I started reading the book out-of-order, thinking that I knew most of this stuff already. But even if you’ve built a few solid web applications already, the book is full of wisdom and very worth reading – after a few chapters I went back to cover-to-cover mode to make sure I wouldn’t miss anything.

Cal’s backdrop is LAMP, but the techniques explained are valid in all environments. I especially like his focus on monitoring, measuring and relentlessy exploring to find out what’s going on inside instead of relying on your gut feelings. We all know that, but some mental hammering of these principles is always good.

Cal’s concrete examples can also help in convincing your colleagues of the benefits of constant measurement and monitoring.

Now, has anyone got a biiiiig system to build so that we can try that in a real setting? ;-)