The path to a file can be thought of as no different from any other text tag. It has one distinction: when combined with sorting within each directory, it imposes a set of total orderings on the collection of files. Pre-order, post-order, breadth-first, each has useful qualities. Hiding the file system erases a host of manifestly useful organizing methods.
You could say that the Dewey decimal system is no better for organizing a library than the ISBN, because either one would locate your book, but that would be false. Arguably any real book ought to appear under a dozen different Dewey numbers, and an encyclopedia under all of them, but somehow that doesn't reduce the practical usefulness of the system.
It would show more insight to demonstrate an understanding of why such artificial, arbitrary systems are so useful, and so improbably usable.
Posted Oct 16, 2012 4:23 UTC (Tue) by Tara_Li (subscriber, #26706)
[Link]
For that matter, how many people actually put useful information in the tags of the pics out of their camera? Maybe, once they put it up on Facebook, do they tag it - but the kind of thing in a photo caption - eg. L to R - Grampa Jimmie Johnson, Grandmaw Kerry Johnson, Grandmaw Shelly Johnson, Uncle Fred, Uncle Thomas - doesn't get added. And then there's the metadata about the metadata... that the Johnsons are a troika marriage...
Ultimately, you spend insane amounts of time putting information about the pics, that you hardly get time to share them with friends.
Plasma Active Three released
Posted Oct 16, 2012 9:38 UTC (Tue) by renox (subscriber, #23785)
[Link]
Tags can be automated, you identify one or two picture and then you have a software tagging all your pictures automatically.
Plasma Active Three released
Posted Oct 16, 2012 9:52 UTC (Tue) by aseigo (guest, #18394)
[Link]
There is definitely a limit on how much detail one will put into tagging. Turns out that's fine: even basic organization does wonders, and it's no more difficult (faster, even) than creating folder hierarchies on disk. So in theory it could suck, but we've found that in practice (with Plasma Active) it doesn't.
Plasma Active Three released
Posted Oct 16, 2012 9:57 UTC (Tue) by aseigo (guest, #18394)
[Link]
If you view a folder path as a kind of tag, then what you say is quite obviously incorrect.
Instead of dragging things into specific folders, drag them into specific tags. Same deal.
And then within each tag (you can view all files that have a specific tag, or set of tags; this is what the UI makes dead simple and fast to do) we can sort however we want.
The functional distinction between a folder name and a tag is really non-existent, except that tags have a lot of additional potential features such as: files can be in multiple tags; tag sets can be used to related data; tags can be applied to non-file data (e.g. people, or rather their contact / identity information), etc.
You mention the Dewey decimal system. You could apply such a system to your files using tags .. in fact, that's precisly what the Dewey system does: it's a simple tagging where everything goes in precisely one tag. This is completely doable with the system presented in Plasma Active .. you just aren't confined to it.
Plasma Active Three released
Posted Oct 16, 2012 13:39 UTC (Tue) by renox (subscriber, #23785)
[Link]
>The functional distinction between a folder name and a tag is really non-existent, except that tags have a lot of additional potential features such as: files can be in multiple tags
This isn't a very good example as with links files can be in multiple directories.
>[cut]tags can be applied to non-file data (e.g. people, or rather their contact / identity information), etc.
In a "Plan9" like organisation, everything is a file so the difference between tags and directory&links would be smaller..
I'm nitpicking, but anyway kudos for using tags! I think that this is a bold move which can be great for the users and I hope that it will succeed, I'm only a bit worried about the performance cost of using tags instead of directories (directories have already a big cost when using HDD as they "hide" the block's position on the disk: http://simula.no/research/nd/publications/Simula.ND.399/s... )
Plasma Active Three released
Posted Oct 16, 2012 19:31 UTC (Tue) by mathstuf (subscriber, #69389)
[Link]
> This isn't a very good example as with links files can be in multiple directories.
Tags are better here. Given a file with multiple links, asking "what other paths point to this inode?" is harder than "what other tags does this file have?".
Implementation detail
Posted Oct 16, 2012 20:42 UTC (Tue) by renox (subscriber, #23785)
[Link]
This depends only on implementation details of links..
This is not the usual way to do it, but I don't see a special difficulty in having the 'link' commands registering the source in an attribute of the destination, in which case you can answer 'what other tags does this file have?' quite easily.
Another way to do it would be to have a separated database which would register such things.
A find command would work too, even if the answer would be much slower.
Implementation detail
Posted Oct 16, 2012 21:21 UTC (Tue) by mathstuf (subscriber, #69389)
[Link]
The problem is that now `mv dir/ foo` is slow if dir/ is large. Every moved file must be checked against the database for multiple links and the new path updated. And we need foo to be fully resolved from the root of the filesystem. Alternatively, each link could be stored in the table as a (name, dirinode) pair and the path generated from the recursion over the dirinode (the mount point that the request was based in should probably be used as the "above the root" path prefix. The functionality could probably also be worked into the locate/updatedb tools.
Implementation detail
Posted Oct 17, 2012 10:32 UTC (Wed) by etienne (subscriber, #25256)
[Link]
> The functionality could probably also be worked into the locate/updatedb tools.
Maybe also most files in the filesystem could be tagged with the name of the rpm/deb package it comes from...
Plasma Active Three released
Posted Oct 18, 2012 7:16 UTC (Thu) by spaetz (subscriber, #32870)
[Link]
There is one argument in the filepath/tag debate I have hardly seen:
Yes, tags might be better and fulfil all needs that filepaths can do, but most tagging systems I know work only within the application you use.
Switch from f-spot to shotwell (or digikam), boom all information is lost. While if it's in the folder 2012/07/Boston/ I have at least some information to reconstruct where the picture was taken. This is why I don't want to give up file hierarchy based locations even if I like tagging.
Plasma Active Three released
Posted Oct 16, 2012 20:26 UTC (Tue) by ncm (subscriber, #165)
[Link]
Aaron, I'm afraid you have so totally missed the point as to have shot off your own foot. I hope it was a result of haste, and not a failure of insight. I was inclined to give you the benefit of the doubt because as you note, you-all are very far from stupid. Tags have proven really useful in organizing photos and e-mail. It's reasonable to speculate that they might serve as well as an overall organizing method for files of all kinds. It's perfectly sensible to knock up a proof-of-concept organizer based on the idea, and publish it so people can try it out.
You write about sorting within a tag, or set of tags. That's not a total ordering, that's a local ordering, and not a stable ordering. Yes, you could model the Dewey system as tags, but such a model would miss most of its value, and anyway would be way too much work for an already otherwise-busy person.
A total ordering is necessarily arbitrary, but it has the virtue of repeatability. A human-animal brain can map its native geographical skills to a repeated presentation, providing an effortless organization that tagging cannot approach. "It was somewhere around here" might mean "perhaps in a folder next to" a known one. What I'm looking for doesn't have tag X; it just showed up in the list a little before something that had tag X.
If tags had natural, and optionally imposed, relationships, that could provide a stable total order. That would correspond, organizationally, to putting folders in other folders. If almost everything that was X was also Y, then those that are X but not Y could automatically be grouped without my asking for "X but not Y".
I would feel a lot more confident about your approach if I saw evidence that you were thinking along these lines. What I'm getting, instead, is that these concepts are totally foreign. Are you old enough to remember when Artificial Intelligence based on Formal Logic was right around the corner? When the Japanese national initiative in support of Logic Programming would rocket Japan ahead of us shlubs? When teaching set theory to pre-schoolers would make arithmetic and algebra, when encountered, intuitively obvious to them all?