Zmarties

For the best possible output from a digital camera, most people capture the data in the camera's RAW format, then process it later. This has the advantage of

• capturing at the camera's full colour depth - which is often much larger than the 8bpp per colour channel that JPEG supports
• no lossy conversion to JPEG at all - so no data is lost
• the ability to adjust the colour temperature of the image
• the ability to apply more computing power to the image

The range of RAW formats is large - even within a camera manufacturer, different models produce different formats of data, and support within image viewing and photo applications is patchy.

I'm pleased to report that IrfanView, one of my favourite image viewers, has just released version 3.97 of formats.dll, which is the plugin for RAW image support. There is no readme to accompany the update, but it does now allow IrfanView to display Minolta RAW files, as produced by the Dynax 7d (aka Maxxum 7d or Alpha 7d). Previously these just caused the program to enter an infinite loop, taking up 100% of the CPU. The bad news is that the support is attrocious! A single preview view of a RAW image takes 8 seconds on my powerful desktop PC, and this is the same time to load the full image - whereupon it is displayed with a horrendous colour cast.

My best recommendation for viewing RAW format files is RawShooter Essentials from Pixmantec. The program has a very strange UI which takes a bit of getting used to (for example it gets rid of the menu bar), but does do the job more than adequately. The unusable 1.0 release was fortunately followed up with a 1.1 release very quickly, which fixed a lot of bugs, and added support for many RAW formats that were claimed for the 1.0 version, but which failed to work reliably.

It's worth noting that Picasa 2 from Google also claims to support RAW images, but it also simply hangs when presented with any examples I've come up with.

Labels: picasa

A new Labs experiment from Google is Google X, which offers a OS X like row of icons which magnify as you mouse over. The icons represent the various sevices Google offers, replacing the standard text links on the home page.

Another of those spare time projects, blogged about on the Google Blog by its creator.

The University of Glasgow has a project called Terrier, which is described as

"software for the rapid development of Web, intranet and desktop search engines. More generally, it is a modular platform for the rapid development of large-scale Information Retrieval applications, providing indexing and retrieval functionalities."

Its a search engine, written in Java, and made available under the Mozilla Public Licence.

Keeping in with the dog naming theme, they also have a linked program called Labrador which is a distributed web crawler (or spider), written in Perl.

Gigablast is not in the big league of search engines, certainly in terms of public awareness and number of queries served, though its index size of 1.5 billion pages is very respectable, up from just 0.5 billion 6 months ago.

In an effort to further fund its expansion, it is looking for venture capital funding, offering 25% of the company for $8.5 million USD. That's a high valuation for a company that until very recently consisted of just Matt Wells, though perhaps there are hidden depths that the website does not show. Certainly Gigablast has some features (such as searching generic metatags) that make it one of my favorite search engines, and it has a refreshingly different approach from the current big boys.

Google have added a Japanese version of Google Suggest, for suggesting matches in real time as you type in a search query.

The data it is driven off is presumably specific to Japanese searchers, so its interesting to see what it returns for each of the initial characters I originally tried with the English language version.

• a - amazon
• b - buffalo
• c - canon
• d - dell
• e - epson
• f -firefox
• g - goo
• h - hotmail
• i - ipod
• j - jal
• k - kakaku
• l - livedoor
• m - msn
• n - nhk
• o - ocn
• p - psp
• q - quicktime
• r - rakuten
• s -sony
• t - tbs
• u - ufj
• v - vector
• w - winny
• x - x live.tv
• y - yahoo
• z - zakzak
• 0 - 0721
• 1 - 1rk
• 2 - 2ch
• 3 - 3m
• 4 - 4gamer
• 5 - 5w1h
• 6 - 603sh
• 7 - 702nk
• 8 - 87
• 9 - 99bb
• . - .htaccess
  

There is a new Customize this Page option appearing on Google News (including international versions).

This offers the ability to define which sections of news appear, and in what order, and how many stories they contain. There is also the ability to add custom sections where the stories contain particular keywords (though all keywords must appear - why no option for an OR search?)

The customization is saved via a cookie - rather unfortunate that since it means the customization is not portable between computers, and is limited to just one set. I'd prefer a additional URL parameter so that I can bookmark different sets of news - the news I want during the week is different from the set I want to read at the weekends, and I may want to share the customizations around a team of people, or access it from different computers. However, I'm glad they have not gone the "sign in to read personalized news" route.

NameVoyager is an interesting way to present a mass of data, using an as you type technique to narrow the data being viewed.

In this case the data is the distribution of the top 1000 US baby names for each of the decades of the last century. The initial display (via a Java applet), shows a mass of lines, representing all the names stacked together, and although some names have wide enough bands to be labelled with the name, most are too thin to do so. Moving the mouse over this highlights a single band, and gives a popup showing the name, and its rank in the decade the mouse is in.

Typing letters narrows the display to just the names matching the typed prefix - with an animated transition as you add additional letters.

Wists provides for visual bookmarks. You select one of the images on a page (via a mechanism involving a bookmarklet which cuts out the cluttering text, to show you just the images), and a thumbnail created from this serves as your bookmark. Like many current web services, it then builds a social experience on top of your lists, and offers tagging and xml feeds.

Its the latest creation from David Galbraith, who was a co founder of Moreover.

Ancient maps would show the known area at some level of detail, but outside that the mapmakers would exercise their imagination - often displaying dragons and other fanciful creatures.

I note that Google Maps has changed its behaviour if you scroll outside the known universe - namely the North American continent. Previously it used to show the wide blue sea stretching ad infinitum, but now it shows an empty grey for these areas instead.

There have been a number of other visual changes:

• the examples on the home page have been changed to be much simpler, and not to confusingly look like search boxes (Peter Merholtz has a screen shot of how it used to look)

Underlying this, the javascript engine driving the UI has been updated to maps.3.js which breaks a number of bookmarklets including my own Latitude and Longitude.

XMLHttpRequest Usability Guidelines is a useful quick summary of how you might use XMLHttpRequest in the UI of a web application.

It both mentions bad uses (such as Google Maps - where the usage breaks the back button and bookmarking experiences), and gives tips on good usage.

Greasemonkey is a Firefox extension that allows you to easily write site specific enhancements.

This has been used to add a persistent searches box for Gmail, and also a Toggle Font link to switch to displaying the current message in an fixed pitch font.

Mihai Parparita, the author of this elegant enhancement to Gmail, happens to work for Google, but has produced this independently of that fact - though you would think that he may be in a reasonable position to get this noticed and rolled into a future Gmail version.