On Mon, 2010-04-05 at 09:44 -0400, Evan Prodromou wrote:
> Giorgos Logiotatidis wrote:
> > Hello everyone,
> >
> > I'm posting my view on the proposed idea "People search server" for the
> > Google Summer of Code [1]. Being able to search for people and groups
> > among different websites supporting the ostatus protocol is a must for
> > the network to grow and evolve as a distributed open source social
> > platform.
> >
> > The search server could read feeds from sites about new registrations,
> > updates, etc.
> It'd probably be better to PuSH them.

Yes that could be a better option. Does statusnet already have this
feature to push registration and profile updates, or this is going to be
part of the GSoC project?

> > Feeds are parsed and saved in a local database by a daemon
> > (besides the RDBMS, a noSQL database can be used for increased speed and
> > better structure of the data). 
> >   
> Really? Are NoSQL databases faster than RDBMSes? Are they better for 
> structuring data?
Not actually having ever worked with NoSQL dbs I don't know if this is
going to be better for statusnet or not but yes there are serious claims
that NoSQL dbs are faster for specific types of applications. But since
you suggest (later on) that this is a statusnet plugin we better follow
the DB model statusnet is already using ;)


> > A Django or PHP powered website can provide to the end user a simple,
> > clean interface to search for people and get a direct answer from our
> > local database. 
> Maybe a smart way to do it would be to build a plugin on top of 
> StatusNet. We've got most of the basics (profile database, OStatus 
> subscription, object model) built-in already.
> In general, the more your proposed codebase diverges from StatusNet's 
> (programming language/libraries/frameworks), the less likely we are to 
> accept it. We'll need to maintain this code in the future.

OK that can be done. I though -reading the description in the status.net
wiki page- that is going to be more 'independent' project. But of course
it can be code as a statusnet plugin

> > Spam bots submitting feeds, or searching, can be easily blocked by
> > integrating recaptcha [2] but more sophisticated techniques would need
> > something better (does spamassasin already have rules for
> > microblogging? :)
> >   
> Probably not.
> > Spoofing will be an interesting field for research. Using GPG keys -for
> > the techies- or a "web-of-trust" -e.g. promoting users that have many
> > subscribers to the search results- could be interesting approaches.
> >   
> Again, PuSH might be the answer here. However, we do already have RSA 
> keys related to accounts, so there may be some value.
> > As an extra: A crawler can be build to automatically find ostatus
> > compatible websites -carefully avoiding spam- and add their user
> > registration feeds to our aggregator.
> >   
> This is an incredibly hard job.
> 
> -Evan
> 

So to clear things out. We need somehow (PuSH sounds nice) to get
registration and profile updates from other ostatus websites, save them
locally and search into that to find people and groups that we are
interested in, while avoiding spammers and spoofers. 

The whole thing should be build as a statusnet plugin, staying as close
as possible to the current provided functionality of statusnet and rules
of it's development, to keep the code base maintainable.

Does this sound like a good GSoC project? 

Thanks for the feedback.

Regards,
Giorgos


_______________________________________________
StatusNet-dev mailing list
[email protected]
http://lists.status.net/mailman/listinfo/statusnet-dev

Reply via email to