The Design of Software (CLOSED)

A public forum for discussing the design of software, from the user interface to the code architecture. Now closed.

The "Design of Software" discussion group has been merged with the main Joel on Software discussion group.

The archives will remain online indefinitely.

Idea for Software - Tag-based PIM free-form Database

Firstly, i'd like to say that im no programmer, just a long time windows user and someone that has to deal with lots of information everyday.

After getting addicted on, i put down on 'paper' my ideas on how great a tag-based PIM would be. Basically, it's all about how outlining/hierarchical software get in the way of the user when it comes down to inputing, sorting and finding information.

You can check it out on my blog:

Unfortunately, there are no software today that meets even half of the my requirements on that paper.

So, what do you guys think on the possibility of a tag-based personal information manager?

Cheers from Brazil,

- Paulo
Paulo Diniz Send private email
Saturday, August 19, 2006
I think the benefit of TAGs is the 'social sharing' benefit: that many people can tag a document.

So, YOU might tag a this form as "programming" and I might tag it as Entreprenuership.  It's taking advantage of the "wisdom of crowds".

Tagging for a single, or small number of users has little benefit because it requires that you "touch" (tag) EVERY DOCUMENT you want to be 'findable'.  I've currently got 51,200 documents on my pc alone indexed. If you spent just 10 seconds on each one, that would be
A FAR BETTER solution is to make it searchable by keyword.

That's 8533 minutes. My time is worth *about* $1 a minute. And that's for doing something interesting.  It's 2x that for boring #@%%.

So, that's a cost of about $17,000. And doesn't guarantee you can FIND the document again.

I'd rather let Copernic Desktop Search do the indexing and I'll do the searching.
Mr. Analogy {Shrinkwrap µISV since 1995} Send private email
Saturday, August 19, 2006
I find most tag implmenentations, like those on highly unintuative.

I have fewer than 100 bookmarks in delicious, but I have probably 30+ tags.  Most of those are orphan tags, which make them useless.  I can "fix" this by going through each bookmark and changing them, but that would take forever and still doesn't keep me from getting new orphans later.

Worse, I can't ask delicious (or flickr, or my own website for that matter) to return sets of tags - "Find me all bookmarks with 'Perl' and 'Vim'".  All I can do is navigate from one broad category to another.  What good is that?  Shouldn't tags let me narrow a search?  Isn't that a basic way of finding information?

Go to and find yourself all the 3.2 volt, 256MB, PC3200 DIMM's.  Then change your mind and decide you want 512MB DIMM's.  Pretty slick, huh?  Can't do that on delicious, flickr, et al, can you?

*That* is how a "tag" system should work.  Take a broad thing like "programming" and narrow it down by other tags - "perl" + "CPAN".

My point?  Dont use delicious as a template for your own system.

PS:  If you take my advice, let me tell you that doing such a system from scratch is hard.  Luckly there is a really awesome tool that will do it all for you; Endeca. Looking at the URL's newegg uses; I can tell you they are using it.  However, Endeca will set you back probably $10k + a dedicated server + god knows what for a support contract.  It's worth it though - it will cost you FAR more to develop what they've done using SQL.
Cory R. King
Saturday, August 19, 2006
For a non-programmer, you did a tremendous job of describing all the various functionalities you would like. When you compare organizing by tree versus organizing by tags, I would agree that tags is better and not yet properly exploited -- it will take a combination of good usability features to get more people interested for personal use. I would agree with the comment that tagging for personal use has a different dynamic than the web tagging you are comparing with; however it is your comparison between tree organization and tag organization that is the key point. I also agree with the comment that google style simple text searching is the key (more than tags) to finding things -- but still organizaing things and finding things are two different parts of the coin. So, all around it is an excellent and well described product idea. I am not too familiar with PIM software though.
Ben Bryant
Saturday, August 19, 2006
I also agree that a tag-based system (rather than a hierarchical folder-based sysem) is much more useful. In fact, I have a pet project I am currently working on. I might release it to the wild at some point.  ;)
Monday, August 21, 2006
It is actually kinda scary. You almost read my mind with your feature list. We do need tin foil hats!
Monday, August 21, 2006
Paolo, your ideas remind me of the Chandler project:
I recommend that you post to their mailing list or Wiki.  They might be able to use some of your ideas.
Monday, August 21, 2006
Thing is, while the folder model is restrictive in some ways (i.e., the "try to remember how you classified this" game) there's information conveyed in the structure that's difficult to replicate cleanly with just tags.  What I really want is a way to associate tags with each other in parent-child relationships, so I can easily roll up all the children of a single parent.  I want a "Sunset Drive-In" tag that knows it's a child of both the Earth:North America:USA:New Mexico:Albuquerque hierarchy *and* the Businesses:Entertainment:Movie Theatres:Drive-Ins hierarchy.
Lazlo Send private email
Monday, August 21, 2006
First off i like to thank all the people in the thread for the interest and also for the critique on my approach!

For those who are interested on a TAG/Based PIM, please note that i have updated a 2.0 version of my essay on my blog.

It is more detailed, has some new ideas, and also various software usage examples. It is much more well written than the first one too. ;-)

Maybe we can get someone (or more) interested on developing such app.

I left the early draft still there for reading. Thanks!

@cory king: Of course you can intersect tags on delicious. check it out with your own eyes all my entries tagged "music" and "portal:

@mranalogy: i certainly agree with your point on time cost, and i'm sure that such system will only be successful if a fast way to feed the database from anywhere on the system (e.g. a system hotkey for filing a new entry) is implemented, to make it second nature to the user, so he can file and tag information as he thinks about it, effortlessly, without disrupting this flow of attention.

Such tagging system is suited for filing and grouping random bits of information and building knowledge on it, because i found it very hard to sort granular and random information using a hierarchical system, and also very time consuming to create a new file in my home folder for every little note i wish to keep, and them having to do a raw search on them all without doing a previous categorization/organization. Like Ben Bryant said above, organizing and finding things are different sides of the coin. With a proper search string, i would find the bit of info i wanted, but i would miss the opportunity to see it side-by-side to another relevant analogous information that are scattered around my computer.

On having to tag information as you file it, it wouldn't be mandatory. If you don't want, that's fine, because you can add it untagged and visualize and search it from the 'global view' (where no tags would be queried). But you would lose the ability to group similar entries.

@outback: If you think we're thinking alike, go to this forum thread:

and read what user "Tom Maisey" have written there.
Very similar to the idea too. I'm "OrcsBR" on that thread.

I'm very looking forward to take a look on your "pet" project. When you are going to let people see it?

@flow: i noticed that too a while ago and sent Mitch Kapor  of OSAF a very early draft of my ideas. He forwarded it to the design list.

Though, Chandler has a different focus, as a traditional PIM software (calendar/mail/tasks/contacts) with some tagging capabilities...  "taglogger" idea is more like a general database.

@lazlo: i have approached this problem in my 2.0 draft. I call it contain/contained relation between tags (item 7.3). It is sort like a parent/child relation that can be estabilished between tags, but with two differences from a hierachical tree: first, like you said, one tag can be child of than one parent tags. Second, and most importantly, i think these relations could be estabilish without relying on the hierarchical metaphor that has created the problem that "Taglogger" is trying to solve. The user is so used to working in a hierarchy, that if one is presented back to him, it would risk returning the user to feeling the URGE to manage a huge structure of metadata relations before deciding which SINGLE tag would be most appropriate to use with the data, which is exactly the problem of using taxonomical classifications on open and general knowledge bases.

- Paulo
Paulo Diniz Send private email
Tuesday, August 22, 2006
Paulo Diniz Send private email
Friday, August 25, 2006
I think it's a good idea that it has tags and full text search. A cloud diagram of the tags is essential. I'm on <cough> dialup at the moment so I haven't checked your blog but I don't understand the "full text not tags" slant come up early on. Both is the key.
Friday, September 01, 2006
Check out the "groups" in the new !ntellect or the older Time & Chaos PIM.  The groups (labels) and categories allow simple but effective organization of PIM type data:
David Burch Send private email
Friday, September 01, 2006

This topic is archived. No further replies will be accepted.

Other recent topics Other recent topics
Powered by FogBugz