Joshua Porter’s workshop at recent UX London conference was great and a must see if you can, but there is something that’s been bugging me since. In discussion about difficulties of getting people to sign-up for service, Joshua mentioned how some companies see connecting with Facebook Connect as Facebook “owning” their users.

They are right, but it doesn’t matter much. Relying on Facebook for authentication certainly makes Facebook sort of a gatekeeper. This does not matter because it can and should be a transitional phase in an evolving relationship.

My view is based on assumption that company wants to build a long lasting relationship with a customer. This may not be true, but then there really is no point in having a sign-up at all. Avoid friction of one, make a trade and move on.

I think most user lifecycle strategies look like trying to get laid by the end of evening while hoping to start a long-term relationship. It might work, but how likely?

There are two things relationships need. Continuing benefit to parties involved and mutual trust. And trust is built through actions over time.

Traditional UX life cycle can be seen as:

Awareness -> Sign-up -> First-time use -> Engagement -> Referral

Each step has to offer more benefits to person taking it as it requires more trust. Moving sign-up after first-time use changes dynamic by making service prove itself before asking same of its user.

Using Facebook Connect (or OAuth , OpenID …) is just going further in mimicking natural relationship building. It’s asking of a smaller commitment at a fragile stage of relationship. You can ask for a bigger commitment with your own authentication later on, when you have already proven your worth and gain enough trust. You could even offer something available only to “full” members to make such commitment more enticing.

If you are building years long relationship like I have with Amazon, does it really matter if you get me to fully sign-up after third month instead of first?

Hence sign-up should be thought of as a gradual process taken over time instead of one-time obstacle.

All this is just speculation at this point, but I hope to test it on two projects this year. If I do, I’ll let you know how it went.

April is about to end and with it also the deadline to submit your EuroPython talk . I thought about submitting a proposal, but realized it would be nuts to do so with my current workload.

It would also be nuts not to go and if you are a European (or not) Python developer you owe it to yourself to come.

I go to a few conferences every year and each has its own personality. It is sometimes hard to point at what makes them (feel) different, but they just are. I like many, but only EuroPython, which I regretfully had to miss last year, feels like home.

That’s why I decided to volunteer this year. It’s a way to give a little bit back, like cleaning dishes when you visit mom.

So do submit your talk and come. It won’t be the same without you.

I admire the look and feel of Mike Matas’ new website . It is really well thought through. I was also intrigued by how he did it, especially after getting a pop-up on my first visit advising me to use a more modern browser than a recent version of Firefox.

There is no point in speculating why some of its features don’t work in more browsers. But I was surprised to see that blurred images are served that way and don’t get blurred in browser. I am playing with an idea of implementing a gallery inspired by Mike’s work, but I would like to reduce manual labor needed for maintaining it.

So I wrote a function that blurs an image on canvas. You can see it in action or copy its code, if you find it useful.

The algorithm used is described in 2001 paper by Wojciech Jarosz. Page contains two implementations, second trading algorithm purity for in my opinion nicer code. Increase number of passes or run it few times over an image, if you need a blurrier result.? Please ask, if you need help with its use.

I also measured its speed to see if it fits my needs. That brought a new surprise. Firefox 3.5.8? on my Linux powered VAIO with 1.2GHz processor blurs image twice as fast as same browser on a Mac with a 2.8Ghz processor. Numbers between runs may vary slightly, but never much. No idea why this is happening, since all functions do is some basic math over items in array that should be well optimized everywhere.

I am sure somebody can optimize it further, but I find it good enough for my use. Image isn’t very blurred after one pass, but one pass over a small image is also a good way to measure how fast a particular computer-browser combination is. On fast combinations I might go for multiple passes over images in view, but fall back to a single pass or no pass on slower systems.