James Archer in The Return of Design raises a complaint about tagging that I share - why don't sites have decent tag delimiters? It's an interesting question, and one that needs to be solved before tagging crosses over into the mainstream. I'm surprised it hasn't been discussed and debated more. Tagging a photo or an article "sanfrancisco" or "newenglandpatriots" just isn't intuitive to the general web user. But there's an even worse problem with the concatenation that's common at services like del.icio.us or Flickr. It results in polluted metadata. In order for community metadata to be valuable to search, we'll need to clean this up. It also encourages the use of simple, short terms rather than more robust concepts or phrases. Today's systems have obviously been designed to sidestep backend technical challenges. Let's make our software smarter to accommodate our users, rather than asking our users to accommodate our software. As Archer says: Programmers and developers are accustomed to oddball text separators, but the general public isn’t. As these services become increasingly widespread, more people are going to be puzzled about the odd symbols in the middle of otherwise logical phrases. Exactly. You know you have a user interface problem when your description of it starts with "programmers and developers are accustomed to..." Space-delimited entry systems are not an option. Yes, consumers are used to the search box, and they know how to type "san francisco museums." But those aren't three distinct tags - "san," "francisco," and "museums" - they're at least three separate overlapping concepts - San Francisco, museums, and San Francisco museums, and all three are relevant. I disagree with the commenter who posted that "Even my mother knows that she needs to use quotes to search for 'a phrase', otherwise she’s just listing a bunch of keywords." Well, your mom isn't a typical search user - the vast majority of search users don't know how to do that. Plus, in my example is "San Francisco" the phrase, or is "San Francisco Museums" the phrase? They both are. In search, the search engine has evolved over time - primarily through the help of people who build web pages - to interpret those concepts. Can't tagging systems get that smart too? What are the UI alternatives? Multiple entry boxes are bad because they make the user do extra work, and they also limit the number of tags to the number of entry fields available in the UI. We want to encourage users to tag to their hearts' content, to use as many tags as are appropriate. Pick-lists or drop-downs are even uglier because they limit the possibilities to the tags provided in the list and make it difficult to extend the set and be ad hoc. Even if the form allows for free-form entry, it implies to the user that there are "standard" or "official" tags that take precedence over others. They also degrade pretty rapidly as the tag set grows even moderately. One thing I've noticed about tagging services is that the ratio of items to tags is very different than the number of items to folders in older systems. My email box contains thousands of emails and eight folders. My del.icio.us list contains 279 URLs and uses 180 tags. Wow. Archer suggests commas for delimiters, and I think that's worth exploring. It's a scenario that's familiar to email users, or at least to email users who send mail to multiple recipients at a time (that's not everyone, by the way). It needs some research, testing, and iteration. There are other UI challenges in tagging for sure. To begin with, what should we call them? We all like tags, but will that make sense to mainstream web users? Should we also look at "keywords," "search terms," "descriptive words," or "labels"? Something else? And how to I jog my memory and prune my personal tag garden (did I use "cycling" or "biking" in the past to describe this concept)? Archer is asking exactly the right questions, and we need more people asking them. When you get excited about an emerging technology it's sometimes easy to tell yourself that eventually the rest of the users will get it. They will, but only after some very smart people think about their needs and design systems and interfaces that make it monkey-simple for them to use. Not because they're stupid, but because they don't have the time nor the energy to fiddle with our obscure technologies.