@reiver

I'm still in the phase of finding out, if there might already be all the tools available, but it seems, I won't get around writing at least some code for it.

I'll probably start extending an existing AP implementation, or using a library, instead of writing everything from scratch. But nothing is set in stone yet. Do you have any suggestions?

@cdonat @reiver my main concern with such a system is it encourages the centralisation and the "big instance" problem, because of the nature of job ads and CV/resume posters. If you're looking for a job, you want your CV federated to as many potential employers as possible, and if you're hiring, you want your job as to be seen as much. I love the idea but without a good relay network it would be difficult to make it cohesive.

@martyn @reiver

I get your concerns. As @reiver already pointed out, the issue could be solved by federated search engines. They can also have federated indices.

Please allow me to respond with a longer thread, to describe my vision. It might change, over time, of course.

1/n

@martyn @reiver

Let's assume I set up a search engine, that uses a hand full of sources known to me, that indexes software developer positions in Switzerland, because that covers broadly, what I'm interested in.

Other search engines as well as AP-Users can then subscribe to a query at my search engine. Whenever my crawler finishes, it publishes the newly found jobs, that match their queries.

2/n

@martyn @reiver

Now the other search engines have a new source. I can also subscribe to their search engines with my search term, and so find job openings, that I wasn't aware of before, because they have other sources.

After I add a job opening from those other search engines to my index, I also match it to my subscribers queries, and push it to them. Since I don't forward jobs, that I already have in my index, this breaks possible loops.

3/n

@martyn @reiver

In the opposite direction, a search engine for recruiters can work similarly. They query known instances for resumes, that match their broad query and then use AP to distribute their results. The federated results can be used by the search engine admin to extend the list of known instances.

4/n

@martyn @reiver

As job seeker, I subscribe to query, that describes the jobs, I'm interested in, and find job openings in my feed.

As recruiter, I subscribe to a query, that describes the candidates for my job openings, and will find matching candidates in my feed.

5/5

@cdonat @reiver I like the idea in principal, do you see those secondary subscribe actions as manual or automatic? I can see pros and cons to both but I feel automatic is the way forward for adoption, perhaps with an option to queue like moderation tools would have for those who don't want it? I'm getting into details here but I feel like a recommendation in the protocol/design doc would be worthwhile. How about multiple CVs for the same person? I have alt. ones for types of jobs for instance.

@martyn @reiver

The idea currently isn't in the state, that I started thinking of that. It's good to discuss all these details, I haven't thought about yet. Thank you for that.

Automatic subscription: I think, we have to distinguish between user, and search engine. The search engines will probably want to add new subscriptions automatically. The user probably wants to only subscribe to his chosen search engine.

@martyn @reiver

Multiple CVs: I thought of the CV as an actor. So having multiple CVs is just having multiple actors on your instance.