These are my answers to two comments regarding the illogical interface:

I would say regarding Joshua's comment, that nobody uses an intersection, because it is *too hard* to use from, and very few people have figured it out! Not having logical operators makes it very difficult to search your own bookmarks, let alone those of others. It is difficult to replace your bookmark folders with tags. Logical operators can of course, substitute for and extend a hierarchy (the "and" operator alone implements a hierarchy). Another question: if there are no spaces in tags, how can someone search on say "Gulf of Mexico"? Adding "of" as a tag totally defeats the purpose of tagging. Semantically, "of" and "Mexico" pretty much are unrelated to "Gulf of Mexico". I see that some people have been using underscores - this isn't really documented anywhere, and is a real hack, to overcome this lack of basic functionality. If semantic tagging is to prove useful, this must be implemented somehow ...

Regarding the tag search on, I think Joshua has a valid point in noting that current users rarely use intersection searches. However, because of the open API, other sites can implement whatever they find lacking, and this may in turn be incorporated into I also think it was a stroke of genious to use only the space character as a tag separator, and leave everything else for the tags. That way, nobody has to learn escape sequences, and every conceivable special character can be used as the user wants. Finally, according to the blog (Update: sorry, the link is dead), it doesn't seem like he's resting on his laurels. Maybe we'll see this in some future version.

About tagging stuff like the "Gulf of Mexico", the solution is the reason for both the popularity and some problems of folksonomies: You get to decide! The best solution for someone living in Northern Norway might be "GulfofMexico", "gulf_of_mexico", or any other fully readable tag, to be easily found again upon browsing the tag list. For someone very familiar with the Gulf of Mexico, "gom", or even "gm", may be good enough, and certainly easy to use for tag searches. AFAIK, doesn't at this moment have any mechanism to overcome the obvious problems with this, but I expect it can be solved with e.g. synonym lists, which users can subscribe to. This way, professors in high energy physics may use their own terms when tagging, and laypeople may search their bookmarks using terms familiar to them due to the synonyms.

I just posted this page to The tags included: logical operators and or not xor exclusive or... now, how could one find a link for "exclusive or" but not "not exclusive"? Or how useful are "and" "or" and "not" as tags? as far as the concept of semantic tagging goes, is a wonderful new meme, but it has just that much further to go. I can forsee that Google perhaps will pick up the concept of tagging and social bookmarking together, and be successful at it, if it isn't implemented by I hear that is very popular in the academic community - papers are categorized by keywords, in bibliographic entries, in sites such as in Entrez PubMed. But how can academics really use this without the full functionality? Of course, I'm sure that you get all the results of the tag query, and then filter them locally. This is extremely inefficient of course, and filtering on the backend of would reduce traffic tremendously.

Most of the time, using underscores, InitCaps, or acronyms is "good enough". Flickr pretends to support spaces in tags, but for my part this has never worked. An idea would be to use the Enter character for a delimiter, but I suspect that would be a problem when accomodating how different platforms encode it (\n, \r, or \r\n), and in that it would mean using text boxes instead of the leaner & meaner text field for input.

Regarding the d.i.i. filtering: It downloads all the bookmarks of the user upon login*, and searches that instead of querying the server. This is convenient for me, for several reasons:

  • It provides an easy way to backup bookmarks on a regular basis (you can access and copy the XML files (Update: sorry, the link is dead) directly).
  • I can filter the whole lot using XSLT.
  • The d.i.i. will work even if goes down or out of service.

This behavior might change later; I'm planning to enable searching all users, and this will mean downloading the results of several searches, and then doing the necessary picking and choosing.

*Only if a correct password has been given, and the local bookmarks file is older than six hours