Benji Weber wrote:
2009/4/1 Ricardo Cornet <rcornet@gmail.com>:
On Wed, Apr 1, 2009 at 3:15 AM, Benji Weber <benji@opensuse.org> wrote:
2009/4/1 Ricardo Cornet <rcornet@gmail.com>:
I tested that page and the results are.. well.. not nice.
Well it's a work in progress. You can also register and start tagging things up & rating things. When there's more user generated content the search can be adjusted to favour that.
The basic problem I found with that is that it count on a high number or technically proficient users. People who will find what they want without using the system at all. And regular people who need the system would either get frustrated because they don't understand what is wrong and give up or make sloppy decisions in tagging or no tagging at all.
In my opinion depending on users is not going to work. Most users are lazy...
Either packagers tag their packages, (which I don't think will happen) or we implemente via software something that read descriptions of packages and generates it own semantic dicctionary.
Well this is kind of what we have already. Indexer looks at package descriptions and .desktop files and tries to infer data where possible.
Well, the inference mechanism needs to be reworked.
Unfortunately there are no real rules for most of the metadata, so for every rule there are hundreds of exceptions. I think the only way for it to work is for people to teach the system.
As I already said, that requires a lot of technical people who would rather do something else. For example, there are a lot of false positives, marking them as "not what I want", would take too much effort.
You can already select which source & binary packages make up an application, and define new applications. Then in future runs the indexer will respect your selections when discovering package updates.
The most elementary keyword have few good matches, Googling the internet is still faster and less error prone. The system as it is, would take too much time and effort to be usable for regular people, who in the end is the only real people who would really need it.
How well the latter is going to work is subject to discussion. But it is easier to make software to work than make people to work. There is a lot of people and a lot of apps.
I'm not sure I agree here. Once it's easy enough people do contribute comments & ratings & tags on many websites nowadays. The key thing is making it easy enough to contribute, and we're not there yet. Not least because we can't even share login credentials with novell infrastracture.
-- Benjamin Weber
Making it easy to help is great. But even if it is easy to contribute, system that depends on learning from the users is to slow in growing up to be effective. Lets say my mom wants an app that helps her to organize her teaching classes, istead of finding and app or getting a "not apps of that kind" response, she will get a long list of false positives. She would think "this thing does not work" and go somewhere frustrated by the percieved stupidity of the system. Telling her to "teach" the system is nonsense, and most people would think the same. Ricardo Cornet. -- To unsubscribe, e-mail: opensuse-project+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-project+help@opensuse.org