Zmarties

The latest (beta) Google Toolbar autolink feature recognises ISBN numbers on web pages, and when asked to by the user, links those numbers to relevant information on that book.

There has been much discussion on the feature, fairly well summarised by Danny Sullivan, though much of the discussion has been far from objective. My one line summary is that consumers like the feature; vocal and unenlightened webpage publishers have objected to it - there are minor tweaks that would make it better for all concerned.

(I was also amused by this categorization of the type of web page authors who fear it!)

One area that I have not seen commented on is that the recognition of ISBN numbers is not very good.

• it only recognises ISBN numbers that either have no puctuation, or exclusively use the "-" character as puctuation - this excludes the very widespread use of the space character to separate the groups of digits in the ISBN. If you look on the title page of many books, you will see that many publishers prefer to print ISBNs using spaces rather than hyphens.
• it only handles 10 digit ISBN numbers, not the 13 digit form that is already printed below the UPC bar code, and which will become the official form in a couple of years, once all systems have have time to migrate to it
  

Examples of ISBN numbers, to try the toolbar out on:

• 0596002416 : just digits
• 0-596-00241-6 : digits separated into groups by hyphens
• 0 596-00241 6 : hyphens between the main groups, but not before the separate check digit
• 0 596 00241 6 : spaces throughout - as often printed
• 9 780596 002411 : 13 digit UPC bar code form (note changed check digit)
  
• 978-0-596-00241-1 : 13 digit ISBN form
• 978 0 596 00241 1 : 13 digit ISBN form using spaces

More info on ISBN numbers can be found at

• ISBN Users Manual
• Transition to 13 digit ISBN

Google have not provided a "publisher opt out", since the addition of links is not automatic, but requires a deliberate "consumer opt in" action. Unfortunately this has led to stupid actions by some publishers, notably Barnes and Noble. They now turn all ISBNs on their pages into links automatically - which gives the stupid circular link problem, where you are viewing a page and it contains links to itself. I'm sure this means that B+N have shot themselves in the foot - people who used to use their site will now stop doing so, because their navigation no longer makes sense.

Thats a shame - B+N search was actually better than Amazon's in that it copes with spaces in ISBNs, whereas Amazon's fails to find anything on the Amazon site - but interesting then offers web search results that point to the given book at competitors sites!

The correct reaction by B+N should be to contact Google, to ensure that the autolink feature can be configured to link to them if thats what the user prefers, much as the mapping autolink can already be configured to use Google Maps or even the competitor Yahoo maps. Google have made it clear that more configuration options are very likely - I hope that this is made totally user configurable so that the autolink technology merely provides the "recognition" half of the task - the "what to link to" can then be user configured, so that solutions like linking ISBNs to your local library are possible.

Note that the autolink also fails to work for any pages loaded using the file protocol. I think Google should fix this - content is content, so why should the protocol used to load the page make a difference.

Update: 5/3/05

A good way to find web pages with ISBNs on them is to do a google search for the phrase "isbn * * * 1|2|3|4|5|6|7|8|9|0|x"

Gmail's What New page is listing two new features:

• sending photos from Picasa via your Gmail account (which Picasa does directly, without firing up your browser)
• a basic HTML mode, to extend support to previously unsupported browsers

The first of these is not so much new, as simply newly documented. The second is a very interesting development - not only is Gmail pushing the boundaries of what we expect from a web based application, but its also making sure that older browsers are not left out. As they point out, you never know what support you are going to find when travelling - all part of their mission to make information available everywhere you happen to be.

Google has introduced a movie: operator which serves two quite distinct purposes:

• searching for movies about a certain subject
• telling you what movies are playing, at what times in a certain locality

As we have unfortunately come to expect with these things, the locality informations only works for US localities, even if you use the operator at one of Google's international sites.

To access the location search you use the syntax "movie:90210" or "movie:Seattle" in your query, whereas to get movies about a certain subject you use the syntax "movie:kong". I'm sure you can spot the problem here - by using the same syntax for two distinct features the service has to guess what you meant, and quite frankly gets it wrong.

Many seasoned travellers, and even just the curious want to know what movies were set in a particular geographical location. I've not found how you find movies about Seattle, nor featuring the starts from TV's 90210. (You can find the obvious Seattle movie by using the search movie:"in seattle", and in fact the use of "in" is quite a good modifier for other places, but will miss a lot of references that mention the place but do not include the word "in").

From another international angle, it would be good if the operator "film:" was recognised as a synonym for "movie:".

Update: If you do "movie:seattle a" then the "a" is a stopword, so is ignored, but does cause the logic to treat this as a subject query, rather than a location query.

A quick round up of a few items to do with image searching:

• Yahoo Images is now claiming over 1.5 billion images in their index
• Also at Yahoo Images, some of the advanced features, such as image size and type can now be accessed by trigger words in your textual query. For example
• walpaper of flowers gets refined via spelling correction into wallpaper of flowers, and thus only returns large images.
• black and white images of new york returns only black and white images
SearchEngineWatch has more examples, though some are incorrect - using bw instead of black and white is not acting as a trigger, merely being used as an extra search term - hence the result was certainly that there were more black and white images in the results, but only as a side effect.

• SEW also notes that Yahoo is inlining images into its main search results when trigger words are used. My research shows that the following are treated as triggers: image, photo, photograph, picture, pic, wallpaper (and often their corresponding plurals). Rather surprisingly, the following words are not used as triggers: jpg, jpeg, gif
  
• MSN image search is powered by Picsearch. Maybe "powered by" is the wrong description - the database of images this makes available is particularly poor - it's much smaller than either Google or Yahoo, and can often get bogged down showing a whole load of images from just one site
  

Without any fanfare, Picasa is now offering build 18.84 for download, instead of the build 18.77 which was originally released. This release is not (yet) being rolled out via Picasa's inbuilt automatic check for updates - if you are running build 18.77 then the update check reports that no updates are available, but if you go to the site and redownload, you will get the new version.

Apparently the new release went out to some beta testers with a readme detailing the changes from the previous build, but there is no such file provided with the download. As far as I can tell, mostly through moderator posts in Picasa's support forums, this build has the following changes:

• a fix for the severe database corruption that occurs when you set Picasa to scan a whole disk for images - the symptoms of this were that after scanning the disk, Picasa would then report that no images were found. Note: This seems to have fixed my particular database corruption, though I have seen reports of some others still having corrupted databases with the new build
• a number of fixes to do with the inability to write CDs
  
• a fix for the deletion of pre-existing pal files
• a fix for an export issue that was causing Gimp and eBay to not like some exported pictures
• addition of 16 bit PSD support
  

Labels: picasa

Version 3 of the Google Toolbar has just been released in beta form, and adds the following new features:

• Spelling correction
• Autolink (which turns certain plain text on the page into links)
• Word Translator

These new features are all implemeted via server side magic - as the updated toolbar privacy page explains, they work by sending the contents of the page you are viewing up to Google. This send is done over https, so the data is encrypted during the transfer. The privacy page actually refers to a dictionary service, which the toolbar does not yet provide - whether this is simply what that are now calling word translator, or whether this refers to another service is unclear.

Despite the close relationship Google now has with Firefox, the toolbar remains IE only, which will disapoint many.

There is a good initial roundup of the features at SearchEngineWatch.

Gmail users over the past few weeks have found they have 50 invites to give away. The various services that sprung up to hand out invites have therefore been flooded with them.

In addition, today I received an invite direct from Google - in a mailing entitled "Gmail Update and Invitation". As well as the invite itself, it outlines the fairly recent changes to the service, including POP access, Gmail notifier, and improved contact handling, and goes on to add

We also wanted to thank you. For showing us your support and for being so patient. And to those who have already signed up for Gmail, thank you for giving it a try and for helping us make it better. Our users are what have made this product great. So whether you're just signing up for your account or you've been with us since the beginning, keep letting us know how we can build you the best email service around.

That's it for now. We hope you like Gmail and will share it with your friends. We've got lots of cool new stuff planned and we can't wait for you to see our work in your Gmail accounts! Stay tuned...


So, after almost a year, they are finally getting around to emailing the people who expressed an interest in the service - I guess that probably means that they are getting pretty close to opening it up to anyone.

Nikon have announced that their COOLPIX 7900, 5900 and 7600 compact digital cameras have face priority auto focus.

When the appropriate portrait mode is selected, the camera automatically detects a person’s face at typical portrait taking distances, then activates autofocus to focus upon the face area. Focus stays on the face, even if the image is recomposed, or the person moves. The face is outlined on the LCD display.

Nikon note that the technique can be fooled - it may not detect a face if the subject is not looking directly at the photographer, or if they are wearing sunglasses. Since a camera has limited processing power, and certainly limited time, these restrictions are perhaps understandable. The technology is based on FaceIt from Identix.

For other (PC) based approaches to detecting faces, The Face Detection Homepage is a good place to start. This includes a list of links to online face detector demos. With more power, and more time to play with, most of these can detect faces with far fewer restrictions than Nikon have, though the latest demo from betaface returned me two matches that were coincident - a case I would have expected to have been checked for and discarded.

IPTC metadata in images, already widely used by professionals, has received a recent boost amongst casual and amateur photographers, since Picasa uses it to store its caption and keyword information.

If you dont wan't to fire up Picasa just to look at this IPTC information, a very useful free utility is IPTCExt. This Windows shell extension allows you to add columns to the Explorer details view, holding the IPTC fields, and to set the fields from the file's properties dialog. The utility has gone a few years without an new version, which illustrates the simplicity and elegance of the solution - the only problem with it that I know of is that it does not handle UTF8 encoded IPTC information, but then again that's common with much IPTC handling software.

In a similar vein, there are other Windows shell extensions for handling image metadata that you may find useful:

• EXIF InfoTip displays EXIF info as you mouse over files in Explorer
• pixvue is an extensive application written only as shell extensions, that handles EXIF, IPTC and XMP metadata
• ViewEXIF is a IE and Firefox extension, that displays EXIF data for images in your browser
• Opanda IExif is a shell extension that handles EXIF, GPS, IPTC and also works in Firefox
  

Update: 2/7/05 I have not kept this bookmark up to date with changes in GMaps, so it no longer works. However I would now recommend using Gmap Extras which is a GreaseMonkey script that adds display of lat/long to Google Maps, and also allows for easy bookmarking by name of locations.

---

The Google Maps UI does not show you your current location, though such information is useful for any number of reasons. It does however provide a "link to this page" link, which does show you the current location in the status bar when you mouse over it.

Thus was born the following bookmarklet, which replaces the "link to this page" text with the actual location of the (centre of) the map. It prefers to show the latitude and longitude, but will show the query string if that's all thats available. To make space for this, it also removes the redundant text links for the Print and Email options - you can still click the icons to get these functions, so nothing is lost. The display is done every two seconds - so may be just slightly behind the map when it changes, and it is not updated dynamically whilst you are dragging the map around, but will catch up soon after you stop. If the link ever reverts to the "link to this page" again, simply rerun the bookmarklet. If the link shows a query string (such as "JFK" or "90210"), then double clicking the map will recentre it, and display latitude and longitude.

This bookmarklet has been tested on Mozilla and IE 6 SR1 - I guess it should work on other browsers such as Firefox and the older version of IE 6. I've tried to make it as robust as possible, but the compact style of bookmarklets does not make it easy to cover every eventuality. It does of course rely somewhat on Google not changing their page markup or link URLs in incompatible ways, but it makes no reference to the existing page's javascript, so should be fairly robust in that respect.

Drag this link to your links bar --> Lat & Long

(I couldn't get IE 6 SR1 to allow me to simply drag and drop the link, but I simply created it by hand and that worked fine.)

A number of other people have been seeing what they can do with bookmarklets on Google Maps, such as:

• Automatic route following bookmark plus others
• weather link
• manual lat/log display
• Map GeoURL

When I reviewed the Ask Jeeves desktop search, I found that it crashed on me far too much to be worth keeping on my machine.

However, Ask Jeeves have had discussions with the Mozilla Foundation about contributing the AJ Desktop Search to Mozilla - either just the indexing technology, or perhaps the whole thing. This open sourcing sounds interesting, and even if it comes to nothing, there are likely to be other ways in which AJ will build extensions and plug-ins for Mozilla / Firefox.

AJ have also been in the news this week for buying Bloglines. Bloglines already works a treat in the Mozilla browser - with a bookmarklet allowing feeds to be easily added. However, I would like to see some even closer browser integration - how about being able to directly bookmark a specific post; and Bloglines is not the complete solution since its online reader can't help me to read internal RSS feeds which are not available via the web, so an online / offline hybrid would be ideal.

Yahoo have released a toolbar for Firefox, as announced on their blog, which provides equivalent functionality to their IE standard toolbar (with the Anti-Spy feature of the enhanced toolbar also expected to be added soon).

There are various buttons that lead to the download - though I note that its obviously different code that provides the buttons on different pages - only on some pages does it check that you explicitly have Firefox which it works on, rather than Netscape or Mozilla which are not supported. The button leads to Yahoo's standard long EULA which if you bother to read it (and few if anyone does) shows how inappropriate it is to this software - it gives you the right to run the software in "object code form" - which can hardly apply since Mozilla based extensions are written in Javascript, and hence distributed in source code form.

The actual extension is signed - well done - this addresses one of the points in Peter Torr's essay "How can I trust Firefox", and will hopefully show other developers the way forward. Updates are planned to add Mozilla and Netscape support. The web pages claim the download is 0.4mb in size, but that's just because they have forgotten to update the text from the IE toolbar pages they are based on - the actual size is much less at 0.06mb.

Google Acquisitions is a list of the various (mostly) companies that Google has bought along the way. Updated just yesterday with a purchase from April 2004 that had been missed, the list now consists of:

1. The Deja archive
2. The assets of Outride
3. Pyra Labs (makers of Blogger)
4. Neotonic Software
5. Applied Semantics
6. Kaltix
7. Sprinks
8. Genius Labs
9. Ignite Logic
10. (A substantial investment in) Baidu
11. Picasa
12. Keyhole

There's speculation that Google is in the running to buy About.com, and ever present guesses that it might consider buying flickr.

Google Maps is a lot of fun, and once again pushes the boundaries of what we now expect can be achieved in a browser based application.

Its main features are:

• a map of the USA (and some of Canada) that can be dragged around, and zoomed into interactively, using mouse or keyboard
  
• the map sizes to fit your window, and all transitions happen without reloading the browser page
  
• local search results can be overlaid on the map
• you can calculate driving instructions between two point, and the route is also overlaid on the map

None of these features are unique, but the package is very neatly put together, and polished.

The underlying techniques used are varied, and there's a good introduction at Mapping Google. These include:

• CSS - the map is built up from a lot of individually positioned small 128 pixel square tiles, which get repositioned as you scroll around
  
• JavaScript - to hold it all together
• XML - processed client side, via XSL, especially used for the search results
  
• transparent png images - for pushpins, speech bubbles, and driving route overlays
• an IFrame - as a back button friendly alternative to XMLHTTPRequest
  

Other useful and informative information on Google Maps can be found at:

• Firefox searchbar for Google Maps
• Slashdot discussion on Mapping Google
• Google Maps and XSL
• Google maps and bookmarks
  

This beta only holds data for the USA. If you want to see a similarly featured, though ultimately less usable system that includes European data, check out Maps24 which provides an interactive map via the use of a Java applet. (Their zoom in from a continent wide view to street level is very impressive the first couple of times you see it!)

Also worth looking at this map of Switzerland that I've mentioned before.

Update: More very informative musings are at:

• Making the back button dance
• Still more fun with maps
• Google Maps Bookmarklets
  

I'm not unique in my dislike of the use of Flash to provide annotations on Flickr. After experiencing similar frustrations, the author of Annotating Images with CSS, came up with a number of techniques that allow image hotspots to be annotated, using just CSS (no Javascript is needed).

The technique works great on CSS2 compliant browsers - which means that Mozilla and Firefox are fine, but IE is not supported, due to its deficient handling of :hover. The examples also make use of transparent pngs, another area in which IE is deficient.

The article provides links to a few other techniques, some of which use Javascript to provide the annotations.

Various sites, in particular Aaron Swartz, are reporting that Google is showing image search results in with the main web results. These results show up to three images at the top of the results page.

However, as with Google Print which censors pages to clients located outside the USA, this feature looks to be only provided to US based IP addresses. Clients making the same request via the google.com website, but who are based in the UK or Germany for example, do not get to see these images.

The way to get round this is to use Google's own proxy service, via their translate service. Thus by asking Google to serve up the page after translation from German to English, you do get to see the page (in a very close approximation to) how the USA sees it. The popular example is a search for pepperoni. Since the page is in English in the first place, the German to English translation should leave the page unchanged - there may be other language pairs that work, that's just the one I tend to use.

(As an aside, why does Google so favour the number three - they show 3 Froogle results, three Google print results, and now three images? Just as I always set the number of results returned to 100 rather than the meagre 10, I want to be able to set the number of results in these special categories to a higher number - if they are going to intrude into my normal results, then let me set the preference to zero when I explicitly dont want them, or to a higher number when I suspect that they will show useful results).

Not much one need say about this - a Firefox extension that times cooking an egg for you.

Whilst Flickr gets much of the press these days, there are other photo sharing sites, as this article in Wired notes.

One such site is smugmug, which to my mind has much to recommend it over Flickr

• it offers results as you type when keyword (tag) searching - giving a list of matching keywords sorted by most populous first
• images are shown using straightforward HTML - none of the nasty Flash presentation of Flickr which breaks the normal right mouse actions I expect to be able to do to images
  
• it picks up IPTC keywords already added to the images
• it has a search box right on the home page

Of course, the sheer number of images in Flickr is greater, helped no doubt by the existence of a free basic account option, whereas smugmug charges $30 per year for its lowest level accounts. However, the quality of images in smugmug seems much higher - it's aimed at the photographer rather than the snapshotter.

I previously gave up using MSN Desktop Search, since it kept triggering my virus detector due to its method of writing extracted email attachments to disk.

Microsoft have acknowledged this was a major design flaw, and have addressed it with a refresh of the tool. The search now streams the data direct to the search filter, rather than writing it to disk - which has the secondary effect giving better performance.

This new way of working relies on the plugin IFilters that MSN Search uses implementing IPersistStream as well as IPersistFile - which the Microsoft provided ones already do, but some third party ones may not. The page on the Channel 9 wiki has more on this.

The MSN blog announcement also notes this update (which does not autoupdate) has the following additional fixes:

• Ability to index Outlook items when it isn’t your default email client
• Clear Deskbar text after performing searches
• Fixed a majority of reported Dr. Watson bugs
• Several bug fixes to improve overall stability & robustness