Filtered by This site

Page 3

Reset

Breaking usability principles for usability

February 10, 2004
3 comments This site

In my "blog about Wikipedia"n:/plog/blogitem-040206-1 I mentioned that one thing I didn't like about Wikipedia is that there are too many links that distract you when you're reading. I prefer to read the text when the inline links aren't underlined like they do it on "susning.nu"n:http://susning.nu/susning.fcgi?action=browse&id=Emacs&oldid=EMACS or "metafilter.com"n:http://www.metafilter.com/mefi/31190.

"Slashdot"n:http://slashdot.org/ disagrees with me. Look at this example on Slashdot. So many links that you don't know whether to read it or to click everything.

The convention of good web design usability is that links should look like links. I.e. blue (purple when visited) and underlined. I still wholeheartedly agree to this, but blog articles are exceptions. Normally in blog articles you want to link certain words to ease surfing. That's a good thing, but I say too many links can draw away attention from what is important.

This is why my new solution supports both underlined and not-underlined. Underlined links for things I urge people to click on and non-underlined for all those things that are just references.

Three examples, where the writing objective is to focus on "Search by Location" but you still want to have a reference link to "Google":

  • With Google you can Search by Location
    Not obvious which one is the most important
  • With "Google"n:http://www.google.com you can "Search by Location"n:http://labs.google.com/location
    Neither link more important then the other
  • With "Google"n:http://www.google.com you can Search by Location
    Perfect! First link for reference, next link more important

This is now a proud feature of my website. It would be interesting to hear what some web design usability experts have to say about this.

Quick URLs for some pages

January 18, 2004
0 comments This site

Supposing I want to send someone a URL to one of the pages of my site. Then if the URL is too long it gets complicated. In emails there's the risk that the long URL gets broken up on many lines. And on SMS there's too much typing. So I've introduced Quick URLs.

For example: https://www.peterbe.com/photos/misc/balloon-flying/ballong_skog_eld.jpg/view becomes https://www.peterbe.com/q-004 Much easier to type in when sending a text, isn't it.

I have to manually select every URL that I want to do this too, but this might change with time. When you enter a Quick URL like the one above you get redirected to the real URL.

About page finally written

January 13, 2004
0 comments This site

This is my first attempt on trying to describe some of the technical mechanisms of this site. I will try to keep it updated as I change the website.

Hopefully I will also write a little something about me, myself and I when time allows.

ESI and Squid and Zope

December 17, 2003
0 comments This site

This can get interesting. ESI, as I've understood it, makes it possible to display on a website "Welcome back Mr. Bengtsson" but have other parts of the page cached through something like Squid. So far I've only used Squid to cache the whole page. From the <html> to the </html> tag.

There is now some development to combine ESI and Squid, what I gather from this blog entry on plope.com I hope to be able to experiment with this here on my site.

Back online

December 10, 2003
0 comments This site

Got the server running again. Long story. Now I'm about to upgrade the web server software to use Zope 2.7.0 and the latest Python.

New feature: RSS feed

October 16, 2003
0 comments This site

RSS feed validator accepts my RSS feed This site now has a RSS feed (version 0.91).

The URL is https://www.peterbe.com/rss.xml

An RSS feed is a XMLish summary of what is being updated on a website. It's standardised so it should be possible to view it from any program or web application.

My recommendation for an RSS feed viewer is Feedreader which is free and works really well. Download it, install it and enter https://www.peterbe.com/rss.xml as a new feed.

Loadtesting this site and compare with static Apache

October 15, 2003
1 comment This site

Disclaimer: These tests are not very scientific so values as they stand are irrelevant.

What is interesting is the difference between the two tests.

Basically what I did was that I sucked the whole www.peterbe.com site (22 files, 111KB) down and served it via Apache on the same machine as the Zope server.

Then I did a benchmark test using using ab with 10 users at 10 concurrent requests. The result is that Apache served the files about 30% faster than Zope.

Complete results

BUT! Bare in mind that the files that Apache served up are static whereas the files Zope served up are dynamic. The content of the Zope request depends on extracting data from a database, sorting, filtering and other various conditional statements. Apache can do none of that unless you use some sort of scripting like PHP, mod_python or mod_perl.

Pages served from Zope are also checked through a comprehensive security mechanism that Apache did not in this case. Every single item (22 of them), when extracted, were with security precautions on the Zope test.

The conclusion from this is: Zope is surprisingly fast considering all the work it needs to do compared to static Apache.