18 April 2009

Idle thoughts I

I hate worrying about money. I hate listening to people, who view earning money as their only goal in life and worship those who are successful at it. It makes me feel inadequate :P

An Amazon-type system would be great for torrenting user-created content. Someone's interests can be determined by which content they rate up or down. Even... it could be used with a tag system. Sorry for switching to 'you' here... if 'you' like something, you give it a rating or just a binary tag, and you select which words best describe it from a list. It isn't just a simple voting system >_<>_<

Anyway the goal of that would be to be able to select a label on a user-created content, and be directed to contents that were labeled the same way by people similar to the ones who created the descriptions for the first content, that's what is important. This is because it would be impossible to create a comprehensive system to give 'true' descriptions to contents, such as in its outstanding production qualities or the in-game behavior of its actors or the perceived philosophy or standards of its creator. A well-designed tagging system is better because it will drift to the correct position, as long as the differences between tags are well-defined and easily spaced in the like memepool or the values or outlooks of people in this life; because even if a word doesn't quite mean what you think it means you can tell what other people think it should mean based on the user-created contents that you have viewed and that it applies to.

Returning to determining interests, lol. A one-sentence tangent must always turn into multiple paragraphs.. -_-' It's important to separate how associations are made. Amazon makes pages really long so they can sell stuff, but not everyone is trying to do that. It would probably be best if recommendations for you in particular, based on what you have rated or even what tags you've placed (would probably be harder, both computation and storage space, to use tags because it's so many more variables than a rating and requires comparing many networks, while you can probably simplify ratings comparison a lot more with averaged out network relationships), are on a completely separate page than normal contents, and are not incorporated into the contents page for every signed-in user. Recommendations on any single content page are based on the average network comparisons based on tags and ratings, independent of who is viewing the page. Then the sparse descriptive tags that are assigned to a user-created content, based on the consistency of previous labeling feedback or 'votes', can be used to isolate a particular quality of the content that the viewer likes.

That was another long 'single sentence' >_< I haven't even referenced torrents at all so far... grr. ANYWAY, all this is complemented by emergent features of torrenting: compilations of material are more successful than single items, because of the overhead of searching for and managing torrents. This also might involve the file system and storage.

I don't know whether it would be better to have the torrent as the thing you search for and rate and download, or whether to create (or is it remove??) an additional abstraction layer by having the file as the basic unit? If it's the file then there might be links to and statistics for a torrent for just that file, as well as to a torrent or torrents that compile this file with other, similar or related user-created contents. Both seem to have their drawbacks; if the torrent is the basic unit, then single files are marginalized and someone may download a torrent batch without caring about the details of the files or those who created them. If the file is the basic unit, then I suppose it becomes harder to show information about the torrents that include this file. Maybe searching for, and providing information about contents can be separate from supporting the torrents social architecture, and the 'layers' could compliment each other, instead of intersecting...

But anyway the idea is that if you know that something is worth watching, you can find a collection, which is a recommendation by someone else, that includes the known contents as well as additional other ones which could be worth watching. It also allows you, for example, to download and manage a compilation by a single creator, as well as support it by continuing to seed. Once again -- the expectation of finding future content worth downloading, and being able to support the needs of others without it being a hassle for yourself, are essential for user-created content to be able to seed effectively.

Torrent management is the other half of this. Being able to designate content that you want to be able to 'watch' for when it needs seeding, based on leeching/seeds ratios and historical consistency of that demand. This means diverse ways of tagging your interest in both seeding and being alerted to new user-created content within a particular category or domain of interest, such as an author. (Might be able to use RSS for this? I don't know, I've never used it myself...) It also means more active, or passive ways to search for needed seeds; such as, torrents that most need seeding that have a particular LABEL (as described before); as well as some means of tracking which files you are able to seed from having them on local storage, and this could be difficult...

only two more things :P OpenID, and incorporating torrents into the OS file browser.

It seems.. that some of the problems that reduce seeding when it's needed, are torrent clutter (minor); adding torrents when you are trying to seed; and moving files within the file hierarchy without breaking active or unopened torrents. It is basically the problems: which files are associated with this torrent; and where can I find them. It's possible that the solution to this could be similar to the thumbnail standard used in the standard OS including the Gnome desktop, that allows many different programs to use the same thumbnails for files and provides backup solutions in case of error... it might also be similar to Google Desktop Search or other things I don't know anything about >_< Anyway now that it's been mentioned I'm sure I, or anyone else would be able to figure it out if necessary, lol.

OpenID could be useful for sites associated with an MMO. Being able to use a central login to identify yourself on other sites; I don't know if it would help to have a one-time authentication that would then allow another site to access information associated with characters on your account or otherwise personalize your identity on that site. It would allow you to login to just that site without worrying about the security of your central account, but it also MIGHT mean having to include a way to manage the permanent authentications associated with your MMO's 'OpenID'-concept login method, so you could remove a key, or site, that is misusing information from your authentication... but anyway. It would mean that you could, for example, offer verifiable character-linked identities on a third-party torrent site, to prevent the (of course) company-sponsored torrent site from being the only one that lets you associate your in-game identity. :P As well as forums and so on..

No comments:

Post a Comment